notebooks/v0_1_prediction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/geopandas/_compat.py:111: UserWarning: The Shapely GEOS version (3.9.1-CAPI-1.14.2) is incompatible with the GEOS version PyGEOS was compiled with (3.10.1-CAPI-1.16.0). Conversions between both will be slow.\n",
      "  warnings.warn(\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n"
     ]
    }
   ],
   "source": [
    "# standard libs\n",
    "import os\n",
    "import sys\n",
    "import logging\n",
    "from functools import partial\n",
    "\n",
    "\n",
    "# project lib\n",
    "PROJECT_SRC_PATH = os.path.join(os.path.abspath(''), '..', 'src')\n",
    "sys.path.append(PROJECT_SRC_PATH)\n",
    "\n",
    "import utils\n",
    "import dataset\n",
    "import visualizations\n",
    "import spatial_autocorrelation\n",
    "from prediction_age import AgePredictor, AgeClassifier, AgePredictorComparison, AgeClassifierComparison\n",
    "from preprocessing import *\n",
    "# external libs\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import geopandas as gpd\n",
    "from shapely import wkt\n",
    "\n",
    "\n",
    "\n",
    "from xgboost import XGBRegressor, XGBClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "logging.basicConfig(format='%(asctime)s | %(levelname)s : %(message)s', level=logging.INFO)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_fr = utils.load_data('france', geo=False)\n",
    "df_fr_sample = utils.sample_cities(df_fr, 0.1)\n",
    "\n",
    "df_nl = utils.load_data('netherlands', geo=False)\n",
    "df_nl_sample = utils.sample_cities(df_nl, 0.1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "gadm_france = pd.read_csv(os.path.join('..', 'metadata', 'FR-LAU-GADM-mapping.csv'))\n",
    "gadm_spain = pd.read_csv(os.path.join('..', 'metadata', 'ES-LAU-GADM-mapping.csv'))\n",
    "gadm_netherlands = pd.read_csv(os.path.join('..', 'metadata', 'NL-LAU-GADM-mapping.csv'))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "'neighborhood'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/indexes/base.py:3621\u001b[0m, in \u001b[0;36mIndex.get_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   3620\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m-> 3621\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_engine\u001b[39m.\u001b[39;49mget_loc(casted_key)\n\u001b[1;32m   3622\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mKeyError\u001b[39;00m \u001b[39mas\u001b[39;00m err:\n",
      "File \u001b[0;32mpandas/_libs/index.pyx:136\u001b[0m, in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/index.pyx:163\u001b[0m, in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/hashtable_class_helper.pxi:5198\u001b[0m, in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/hashtable_class_helper.pxi:5206\u001b[0m, in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mKeyError\u001b[0m: 'neighborhood'",
      "\nThe above exception was the direct cause of the following exception:\n",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb Cell 7'\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000087?line=0'>1</a>\u001b[0m df_fr_test \u001b[39m=\u001b[39m pd\u001b[39m.\u001b[39mread_pickle(\u001b[39m'\u001b[39m\u001b[39m../data/prepared/df-FRA.pkl\u001b[39m\u001b[39m'\u001b[39m)\n\u001b[0;32m----> <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000087?line=1'>2</a>\u001b[0m df_fr_test[\u001b[39m'\u001b[39;49m\u001b[39mneighborhood\u001b[39;49m\u001b[39m'\u001b[39;49m]\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/frame.py:3505\u001b[0m, in \u001b[0;36mDataFrame.__getitem__\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   3503\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mcolumns\u001b[39m.\u001b[39mnlevels \u001b[39m>\u001b[39m \u001b[39m1\u001b[39m:\n\u001b[1;32m   3504\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_getitem_multilevel(key)\n\u001b[0;32m-> 3505\u001b[0m indexer \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mcolumns\u001b[39m.\u001b[39;49mget_loc(key)\n\u001b[1;32m   3506\u001b[0m \u001b[39mif\u001b[39;00m is_integer(indexer):\n\u001b[1;32m   3507\u001b[0m     indexer \u001b[39m=\u001b[39m [indexer]\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/indexes/base.py:3623\u001b[0m, in \u001b[0;36mIndex.get_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   3621\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_engine\u001b[39m.\u001b[39mget_loc(casted_key)\n\u001b[1;32m   3622\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mKeyError\u001b[39;00m \u001b[39mas\u001b[39;00m err:\n\u001b[0;32m-> 3623\u001b[0m     \u001b[39mraise\u001b[39;00m \u001b[39mKeyError\u001b[39;00m(key) \u001b[39mfrom\u001b[39;00m \u001b[39merr\u001b[39;00m\n\u001b[1;32m   3624\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mTypeError\u001b[39;00m:\n\u001b[1;32m   3625\u001b[0m     \u001b[39m# If we have a listlike key, _check_indexing_error will raise\u001b[39;00m\n\u001b[1;32m   3626\u001b[0m     \u001b[39m#  InvalidIndexError. Otherwise we fall through and re-raise\u001b[39;00m\n\u001b[1;32m   3627\u001b[0m     \u001b[39m#  the TypeError.\u001b[39;00m\n\u001b[1;32m   3628\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_check_indexing_error(key)\n",
      "\u001b[0;31mKeyError\u001b[0m: 'neighborhood'"
     ]
    }
   ],
   "source": [
    "df_fr_test = pd.read_pickle('../data/prepared/df-FRA.pkl')\n",
    "df_fr_test['neighborhood']\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Sample data for local test runs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_nl_pre = pd.read_pickle('../data/prepared/df-NLD-preliminary.pkl')\n",
    "df_fr_pre = pd.read_pickle('../data/prepared/df-FRA-preliminary.pkl')\n",
    "df_es_pre = pd.read_pickle('../data/prepared/df-ESP-preliminary.pkl')\n",
    "\n",
    "# df_nl_pre_sample = utils.sample_cities(df_nl_pre, n=10)\n",
    "# df_fr_pre_sample = utils.sample_cities(df_fr_pre, n=10)\n",
    "# df_es_pre_sample = utils.sample_cities(df_es_pre, n=10)\n",
    "# df_pre_sample = pd.concat([df_nl_pre_sample, df_fr_pre_sample, df_es_pre_sample], axis=0)\n",
    "\n",
    "# df_nl_pre_sample.to_pickle('../data/exp/df-NLD-exp.pkl')\n",
    "# df_es_pre_sample.to_pickle('../data/exp/df-ESP-exp.pkl')\n",
    "# df_fr_pre_sample.to_pickle('../data/exp/df-FRA-exp.pkl')\n",
    "# df_pre_sample.to_pickle('../data/exp/df-NLD-FRA-ESP-exp.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1319,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[autoreload of prediction_age failed: Traceback (most recent call last):\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 257, in check\n",
      "    superreload(m, reload, self.old_objects)\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 480, in superreload\n",
      "    update_generic(old_obj, new_obj)\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 377, in update_generic\n",
      "    update(a, b)\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 329, in update_class\n",
      "    if update_generic(old_obj, new_obj):\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 377, in update_generic\n",
      "    update(a, b)\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 289, in update_function\n",
      "    setattr(old, name, getattr(new, name))\n",
      "ValueError: mcc() requires a code object with 0 free vars, not 1\n",
      "]\n"
     ]
    }
   ],
   "source": [
    "df_nl_sample = pd.read_pickle('../data/exp/df-NLD-exp.pkl')\n",
    "df_fr_sample = pd.read_pickle('../data/exp/df-FRA-exp.pkl')\n",
    "df_es_sample = pd.read_pickle('../data/exp/df-ESP-exp.pkl')\n",
    "# df_sample = pd.read_pickle('../data/exp/df-NLD-FRA-ESP-exp.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "32114"
      ]
     },
     "execution_count": 114,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# len(df_es_pre.city.unique())\n",
    "# utils.sample_cities(df_es_pre, n=10).groupby('city').count()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 264,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "175298"
      ]
     },
     "execution_count": 264,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# df_fr_pre_res = preparation.add_residential_type_column(df_fr_pre)\n",
    "# df_fr_pre_res[['id', 'country', 'type', 'residential_type', 'floors', 'n_neighbors', 'TouchesIndexes', 'FootprintArea', 'height', 'lat', 'lon']]\n",
    "# import ast\n",
    "# df_nl_pre.loc[0, 'TouchesIndexes'] = np.nan\n",
    "# df_nl_pre['TouchesIndexes'].fillna('[1]').apply(lambda l: len(ast.literal_eval(l)))\n",
    "df_nl_pre['TouchesIndexes'].nunique(dropna=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocessing & Validation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Select cities for generalization experiment with increasing spatial distance between train & test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 262,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['GID_0', 'NAME_0', 'GID_1', 'NAME_1', 'GID_2', 'NAME_2', 'GID_3',\n",
       "       'NAME_3', 'GID_4', 'NAME_4', 'VARNAME_4', 'TYPE_4', 'ENGTYPE_4', 'CC_4',\n",
       "       'geometry', 'index_right', 'LAU_ID', 'LAU_NAME', 'geometry_lau',\n",
       "       'sjoin_distance', 'centroid_distance', 'LAU CODE', 'POPULATION',\n",
       "       'DEGURBA'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 262,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gadm_spain.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 640,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "eu_regions = {}\n",
    "for country, gadm, city_lvl, region_lvl in [('ESP', gadm_spain, 4, 2), ('FRA', gadm_france, 4, 2), ('NLD', gadm_netherlands, 2, 1)]:\n",
    "    eu_regions[country] = gadm.groupby(f'NAME_{region_lvl}')[f'NAME_{city_lvl}'].apply(list).to_dict()\n",
    "\n",
    "with open(os.path.join('..', 'metadata', f'regions.json'), 'w', encoding='utf8') as f:\n",
    "    json.dump(eu_regions, f, indent=2, ensure_ascii=False)\n",
    "\n",
    "eu_large_regions = {}\n",
    "for country, gadm, city_lvl, region_lvl in [('ESP', gadm_spain, 4, 1), ('FRA', gadm_france, 4, 1), ('NLD', gadm_netherlands, 2, 1)]:\n",
    "    eu_large_regions[country] = gadm.groupby(f'NAME_{region_lvl}')[f'NAME_{city_lvl}'].apply(list).to_dict()\n",
    "\n",
    "with open(os.path.join('..', 'metadata', f'large-regions.json'), 'w', encoding='utf8') as f:\n",
    "    json.dump(eu_large_regions, f, indent=2, ensure_ascii=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4091.71293272371"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import json\n",
    "with open(os.path.join('..', 'metadata', f'buildings-per-region-FRA.json'), 'r', encoding='utf-8') as f:\n",
    "    buildings_per_city = json.load(f)\n",
    "np.mean(list(buildings_per_city.values()))\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-09 16:41:08,954 | INFO : Regions dissolved.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "24 regions fullfiled the constraints.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-09 16:41:16,752 | INFO : Regions dissolved.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "55 regions fullfiled the constraints.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-09 16:41:18,296 | INFO : Regions dissolved.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9 regions fullfiled the constraints.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'\\nSpecifically, we use each of the 12 regions of the Netherlands for which we have available building data once as test set and divide buildings from other regions into 6 groups depending on their distance to the test region.\\nThe closest group of cities is at least 25km and at most 50km away from the test region. For every group the distance increases by 25km, so that cities from the sixth group are between 150km and 175km away from the test set.\\nTo make models trained on different groups comparable, we consistently utilize around 150,000 buildings from randomly selected cities within each group for training.\\n'"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# -> create cross validation method from this which enforces a spatial distance between test and train set\n",
    "import json\n",
    "\n",
    "def sample_cities_until_n_buildings(cities, min_n_buildings, buildings_per_city, tolerance=0.3):\n",
    "    random.seed(dataset.GLOBAL_REPRODUCIBILITY_SEED)\n",
    "\n",
    "    for n in range(len(cities)):\n",
    "        sampled_cities = random.sample(cities, n+1)\n",
    "        n_buildings = sum(buildings_per_city.get(city, 0) for city in sampled_cities)\n",
    "\n",
    "        if n_buildings < min_n_buildings:\n",
    "            previous_cities = sampled_cities\n",
    "\n",
    "        elif (n_buildings - min_n_buildings) / min_n_buildings < tolerance:\n",
    "            return sampled_cities\n",
    "\n",
    "        else:\n",
    "            for i in range(300):\n",
    "                random.seed(i)\n",
    "                remaining_cities = [city for city in cities if city not in previous_cities]\n",
    "                sampled_cities = previous_cities + random.sample(remaining_cities, 1)\n",
    "                n_buildings = sum(buildings_per_city.get(city, 0) for city in sampled_cities)\n",
    "\n",
    "                if n_buildings < min_n_buildings:\n",
    "                    previous_cities = sampled_cities\n",
    "\n",
    "                elif (n_buildings - min_n_buildings) / min_n_buildings < tolerance:\n",
    "                    return sampled_cities\n",
    "\n",
    "            raise Exception(f'Sampling cities with the constraint of at least {min_n_buildings} buildings but at most {min_n_buildings * (1 + tolerance)} failed after 50 tries with different seeds with {n_buildings} buildings.')\n",
    "    return []\n",
    "\n",
    "\n",
    "for country, gadm, city_lvl, region_lvl, max_distance in [('ESP', gadm_spain, 4, 2, 500000), ('FRA', gadm_france, 4, 2, 500000), ('NLD', gadm_netherlands, 2, 1, 250000)]:\n",
    "# for country, gadm, city_lvl, region_lvl in [('ESP', gadm_spain, 4, 2)]:\n",
    "\n",
    "    with open(os.path.join('..', 'metadata', f'buildings-per-region-{country}.json'), 'r', encoding='utf-8') as f:\n",
    "        buildings_per_city = json.load(f)\n",
    "    # buildings_per_city = df_nl_full.groupby('city').size().to_dict()\n",
    "\n",
    "    gadm = gadm.dropna(subset=[f'NAME_{city_lvl}'])\n",
    "    gadm_region_lvl = geometry.to_gdf(gadm).dissolve(by=f'GID_{region_lvl}', as_index=False)\n",
    "    # %matplotlib qt\n",
    "    # gadm_region_lvl.plot(column=f'NAME_{region_lvl}', cmap='tab20')\n",
    "    logger.info('Regions dissolved.')\n",
    "\n",
    "    test_regions = gadm_region_lvl\n",
    "    test_region_centroids = test_regions.centroid\n",
    "\n",
    "    matrix = geometry.pairwise_distance_matrix(test_region_centroids, geometry.to_gdf(gadm).centroid)\n",
    "    test_region_names = list(test_regions[f'NAME_{region_lvl}'].values)\n",
    "    train_city_names = list(gadm[f'NAME_{city_lvl}'].values)\n",
    "    city_to_region_mapping = dict(zip(gadm[f'NAME_{city_lvl}'], gadm[f'NAME_{region_lvl}']))\n",
    "    region_to_city_mapping = gadm.groupby(f'NAME_{region_lvl}')[f'NAME_{city_lvl}'].apply(list).to_dict()\n",
    "\n",
    "\n",
    "    import collections\n",
    "    train_test_sets = collections.defaultdict(list)\n",
    "    train_test_sets_stats = collections.defaultdict(list)\n",
    "    for i, neighboring_regions in enumerate(matrix):\n",
    "        test_cities = region_to_city_mapping[test_region_names[i]]\n",
    "        for min in range(50000, max_distance, 50000):\n",
    "            train_cities = [train_city_names[idx] for idx, dis in enumerate(neighboring_regions) if dis > min and dis < min + 50000 and city_to_region_mapping[train_city_names[idx]] != test_region_names[i]]\n",
    "            train_cities = sample_cities_until_n_buildings(train_cities, 100000, buildings_per_city)\n",
    "            n_buildings_train = sum(buildings_per_city.get(city, 0) for city in train_cities)\n",
    "            n_buildings_test = sum(buildings_per_city.get(city, 0) for city in test_cities)\n",
    "\n",
    "            if n_buildings_train < 100000 or n_buildings_test < 100000:  # ESP\n",
    "                train_test_sets[test_region_names[i]] = None\n",
    "                break\n",
    "\n",
    "            train_test_sets[test_region_names[i]].append((train_cities, test_cities))\n",
    "            train_test_sets_stats[test_region_names[i]].append((n_buildings_train, n_buildings_test))\n",
    "\n",
    "    train_test_sets = {k: v for k, v in train_test_sets.items() if v is not None}\n",
    "    print(f'{len(train_test_sets.keys())} regions fullfiled the constraints.')\n",
    "    # print(train_test_sets_stats)\n",
    "\n",
    "    with open(os.path.join('..', 'metadata', f'experiment-spatial-distance-cities-{country}.json'), 'w', encoding='utf8') as f:\n",
    "        json.dump(train_test_sets, f, sort_keys=True, indent=2, ensure_ascii=False)\n",
    "\n",
    "\"\"\"\n",
    "Specifically, we use each of the 12 regions of the Netherlands for which we have available building data once as test set and divide buildings from other regions into 6 groups depending on their distance to the test region.\n",
    "The closest group of cities is at least 25km and at most 50km away from the test region. For every group the distance increases by 25km, so that cities from the sixth group are between 150km and 175km away from the test set.\n",
    "To make models trained on different groups comparable, we consistently utilize around 150,000 buildings from randomly selected cities within each group for training.\n",
    "\"\"\"\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Age bins (harmonization across countries)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Country</th>\n",
       "      <th>BuildingType</th>\n",
       "      <th>Age</th>\n",
       "      <th>Age_Harmonized</th>\n",
       "      <th>q_h_nd</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>187</th>\n",
       "      <td>Spain</td>\n",
       "      <td>AB</td>\n",
       "      <td>1901</td>\n",
       "      <td>1900-1944</td>\n",
       "      <td>11.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>188</th>\n",
       "      <td>Spain</td>\n",
       "      <td>AB</td>\n",
       "      <td>1901-1936</td>\n",
       "      <td>1900-1944</td>\n",
       "      <td>12.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>189</th>\n",
       "      <td>Spain</td>\n",
       "      <td>AB</td>\n",
       "      <td>1937-1959</td>\n",
       "      <td>1945-1969</td>\n",
       "      <td>10.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>190</th>\n",
       "      <td>Spain</td>\n",
       "      <td>AB</td>\n",
       "      <td>1960-1979</td>\n",
       "      <td>1970-1979</td>\n",
       "      <td>5.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>191</th>\n",
       "      <td>Spain</td>\n",
       "      <td>AB</td>\n",
       "      <td>1980-2006</td>\n",
       "      <td>1990-1999</td>\n",
       "      <td>3.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>192</th>\n",
       "      <td>Spain</td>\n",
       "      <td>AB</td>\n",
       "      <td>2007</td>\n",
       "      <td>2000-2009</td>\n",
       "      <td>2.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>193</th>\n",
       "      <td>Spain</td>\n",
       "      <td>MFH</td>\n",
       "      <td>1901</td>\n",
       "      <td>1900-1944</td>\n",
       "      <td>17.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>194</th>\n",
       "      <td>Spain</td>\n",
       "      <td>MFH</td>\n",
       "      <td>1901-1936</td>\n",
       "      <td>1900-1944</td>\n",
       "      <td>10.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>195</th>\n",
       "      <td>Spain</td>\n",
       "      <td>MFH</td>\n",
       "      <td>1937-1959</td>\n",
       "      <td>1945-1969</td>\n",
       "      <td>16.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>196</th>\n",
       "      <td>Spain</td>\n",
       "      <td>MFH</td>\n",
       "      <td>1960-1979</td>\n",
       "      <td>1970-1979</td>\n",
       "      <td>14.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>197</th>\n",
       "      <td>Spain</td>\n",
       "      <td>MFH</td>\n",
       "      <td>1980-2006</td>\n",
       "      <td>1990-1999</td>\n",
       "      <td>5.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>198</th>\n",
       "      <td>Spain</td>\n",
       "      <td>MFH</td>\n",
       "      <td>2007</td>\n",
       "      <td>2000-2009</td>\n",
       "      <td>4.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>199</th>\n",
       "      <td>Spain</td>\n",
       "      <td>SFH</td>\n",
       "      <td>1901</td>\n",
       "      <td>1900-1944</td>\n",
       "      <td>14.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>200</th>\n",
       "      <td>Spain</td>\n",
       "      <td>SFH</td>\n",
       "      <td>1901-1936</td>\n",
       "      <td>1900-1944</td>\n",
       "      <td>21.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201</th>\n",
       "      <td>Spain</td>\n",
       "      <td>SFH</td>\n",
       "      <td>1937-1959</td>\n",
       "      <td>1945-1969</td>\n",
       "      <td>11.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>202</th>\n",
       "      <td>Spain</td>\n",
       "      <td>SFH</td>\n",
       "      <td>1960-1979</td>\n",
       "      <td>1970-1979</td>\n",
       "      <td>17.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>203</th>\n",
       "      <td>Spain</td>\n",
       "      <td>SFH</td>\n",
       "      <td>1980-2006</td>\n",
       "      <td>1990-1999</td>\n",
       "      <td>7.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>204</th>\n",
       "      <td>Spain</td>\n",
       "      <td>SFH</td>\n",
       "      <td>2007</td>\n",
       "      <td>2000-2009</td>\n",
       "      <td>7.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>205</th>\n",
       "      <td>Spain</td>\n",
       "      <td>TH</td>\n",
       "      <td>1901</td>\n",
       "      <td>1900-1944</td>\n",
       "      <td>9.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>206</th>\n",
       "      <td>Spain</td>\n",
       "      <td>TH</td>\n",
       "      <td>1901-1936</td>\n",
       "      <td>1900-1944</td>\n",
       "      <td>27.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>207</th>\n",
       "      <td>Spain</td>\n",
       "      <td>TH</td>\n",
       "      <td>1937-1959</td>\n",
       "      <td>1945-1969</td>\n",
       "      <td>33.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>208</th>\n",
       "      <td>Spain</td>\n",
       "      <td>TH</td>\n",
       "      <td>1960-1979</td>\n",
       "      <td>1970-1979</td>\n",
       "      <td>10.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>209</th>\n",
       "      <td>Spain</td>\n",
       "      <td>TH</td>\n",
       "      <td>1980-2006</td>\n",
       "      <td>1990-1999</td>\n",
       "      <td>7.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>210</th>\n",
       "      <td>Spain</td>\n",
       "      <td>TH</td>\n",
       "      <td>2007</td>\n",
       "      <td>2000-2009</td>\n",
       "      <td>3.1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Country BuildingType        Age Age_Harmonized  q_h_nd\n",
       "187   Spain           AB       1901      1900-1944    11.2\n",
       "188   Spain           AB  1901-1936      1900-1944    12.2\n",
       "189   Spain           AB  1937-1959      1945-1969    10.5\n",
       "190   Spain           AB  1960-1979      1970-1979     5.9\n",
       "191   Spain           AB  1980-2006      1990-1999     3.1\n",
       "192   Spain           AB       2007      2000-2009     2.9\n",
       "193   Spain          MFH       1901      1900-1944    17.6\n",
       "194   Spain          MFH  1901-1936      1900-1944    10.9\n",
       "195   Spain          MFH  1937-1959      1945-1969    16.9\n",
       "196   Spain          MFH  1960-1979      1970-1979    14.3\n",
       "197   Spain          MFH  1980-2006      1990-1999     5.3\n",
       "198   Spain          MFH       2007      2000-2009     4.4\n",
       "199   Spain          SFH       1901      1900-1944    14.8\n",
       "200   Spain          SFH  1901-1936      1900-1944    21.9\n",
       "201   Spain          SFH  1937-1959      1945-1969    11.0\n",
       "202   Spain          SFH  1960-1979      1970-1979    17.7\n",
       "203   Spain          SFH  1980-2006      1990-1999     7.8\n",
       "204   Spain          SFH       2007      2000-2009     7.9\n",
       "205   Spain           TH       1901      1900-1944     9.6\n",
       "206   Spain           TH  1901-1936      1900-1944    27.6\n",
       "207   Spain           TH  1937-1959      1945-1969    33.5\n",
       "208   Spain           TH  1960-1979      1970-1979    10.4\n",
       "209   Spain           TH  1980-2006      1990-1999     7.8\n",
       "210   Spain           TH       2007      2000-2009     3.1"
      ]
     },
     "execution_count": 203,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tabula_bins_path = os.path.join('..', 'metadata', 'TABULA_parameters_harmonized.csv')\n",
    "tabula_bins = pd.read_csv(tabula_bins_path)\n",
    "tabula_bins = tabula_bins[tabula_bins['Country'].isin(['Netherlands', 'Spain', 'France'])]\n",
    "tabula_bins[tabula_bins['Country'] == 'Spain'][['Country', 'BuildingType', 'Age', 'Age_Harmonized', 'q_h_nd']][:30]\n",
    "tabula_energy_path = os.path.join('..', 'metadata', 'TABULA_heating_demand.csv')\n",
    "pd.read_csv(tabula_energy_path) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Residential type"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_fr_pre_res = preparation.add_residential_type_column(df_fr_pre)\n",
    "df_fr_pre_res[['id', 'country', 'type', 'residential_type', 'floors', 'n_neighbors', 'TouchesIndexes', 'FootprintArea', 'height', 'lat', 'lon']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Neighborhood column"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>city</th>\n",
       "      <th>sbb</th>\n",
       "      <th>neighborhood</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.14.7_1-23804</th>\n",
       "      <td>v0.1-NLD.14.7_1-23804</td>\n",
       "      <td>Alphen aan den Rijn</td>\n",
       "      <td>f2d955e7</td>\n",
       "      <td>179c6287</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.14.7_1-23812</th>\n",
       "      <td>v0.1-NLD.14.7_1-23812</td>\n",
       "      <td>Alphen aan den Rijn</td>\n",
       "      <td>cc419c3a</td>\n",
       "      <td>35765232</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.14.7_1-23814</th>\n",
       "      <td>v0.1-NLD.14.7_1-23814</td>\n",
       "      <td>Alphen aan den Rijn</td>\n",
       "      <td>f344cba1</td>\n",
       "      <td>35765232</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.14.7_1-23815</th>\n",
       "      <td>v0.1-NLD.14.7_1-23815</td>\n",
       "      <td>Alphen aan den Rijn</td>\n",
       "      <td>5c8dba04</td>\n",
       "      <td>feb9c6a8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.14.7_1-23817</th>\n",
       "      <td>v0.1-NLD.14.7_1-23817</td>\n",
       "      <td>Alphen aan den Rijn</td>\n",
       "      <td>2cec158f</td>\n",
       "      <td>feb9c6a8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.9.65_1-6398</th>\n",
       "      <td>v0.1-NLD.9.65_1-6398</td>\n",
       "      <td>Zijpe</td>\n",
       "      <td>4e2219d4</td>\n",
       "      <td>0dbe8b5f</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.9.65_1-6399</th>\n",
       "      <td>v0.1-NLD.9.65_1-6399</td>\n",
       "      <td>Zijpe</td>\n",
       "      <td>2e2099b2</td>\n",
       "      <td>0dbe8b5f</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.9.65_1-6400</th>\n",
       "      <td>v0.1-NLD.9.65_1-6400</td>\n",
       "      <td>Zijpe</td>\n",
       "      <td>229962f3</td>\n",
       "      <td>853f0bde</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.9.65_1-6401</th>\n",
       "      <td>v0.1-NLD.9.65_1-6401</td>\n",
       "      <td>Zijpe</td>\n",
       "      <td>229962f3</td>\n",
       "      <td>853f0bde</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-NLD.9.65_1-6402</th>\n",
       "      <td>v0.1-NLD.9.65_1-6402</td>\n",
       "      <td>Zijpe</td>\n",
       "      <td>2e2099b2</td>\n",
       "      <td>0dbe8b5f</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>691981 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                          id                 city       sbb  \\\n",
       "id                                                                            \n",
       "v0.1-NLD.14.7_1-23804  v0.1-NLD.14.7_1-23804  Alphen aan den Rijn  f2d955e7   \n",
       "v0.1-NLD.14.7_1-23812  v0.1-NLD.14.7_1-23812  Alphen aan den Rijn  cc419c3a   \n",
       "v0.1-NLD.14.7_1-23814  v0.1-NLD.14.7_1-23814  Alphen aan den Rijn  f344cba1   \n",
       "v0.1-NLD.14.7_1-23815  v0.1-NLD.14.7_1-23815  Alphen aan den Rijn  5c8dba04   \n",
       "v0.1-NLD.14.7_1-23817  v0.1-NLD.14.7_1-23817  Alphen aan den Rijn  2cec158f   \n",
       "...                                      ...                  ...       ...   \n",
       "v0.1-NLD.9.65_1-6398    v0.1-NLD.9.65_1-6398                Zijpe  4e2219d4   \n",
       "v0.1-NLD.9.65_1-6399    v0.1-NLD.9.65_1-6399                Zijpe  2e2099b2   \n",
       "v0.1-NLD.9.65_1-6400    v0.1-NLD.9.65_1-6400                Zijpe  229962f3   \n",
       "v0.1-NLD.9.65_1-6401    v0.1-NLD.9.65_1-6401                Zijpe  229962f3   \n",
       "v0.1-NLD.9.65_1-6402    v0.1-NLD.9.65_1-6402                Zijpe  2e2099b2   \n",
       "\n",
       "                      neighborhood  \n",
       "id                                  \n",
       "v0.1-NLD.14.7_1-23804     179c6287  \n",
       "v0.1-NLD.14.7_1-23812     35765232  \n",
       "v0.1-NLD.14.7_1-23814     35765232  \n",
       "v0.1-NLD.14.7_1-23815     feb9c6a8  \n",
       "v0.1-NLD.14.7_1-23817     feb9c6a8  \n",
       "...                            ...  \n",
       "v0.1-NLD.9.65_1-6398      0dbe8b5f  \n",
       "v0.1-NLD.9.65_1-6399      0dbe8b5f  \n",
       "v0.1-NLD.9.65_1-6400      853f0bde  \n",
       "v0.1-NLD.9.65_1-6401      853f0bde  \n",
       "v0.1-NLD.9.65_1-6402      0dbe8b5f  \n",
       "\n",
       "[691981 rows x 4 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_nl_pre[['id', 'city', 'sbb', 'neighborhood']].set_index('id', drop=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Experiments"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Preliminary experiments"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Feature importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "27 of 120 features are significant for regression.\n",
      "31 of 120 features are significant for classification.\n",
      "0.19 of features are significant for both regression and classification.\n",
      "0.6571428571428571 of the significant features are significant for both regression and classification.\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x864 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fts_importance_reg = pd.read_csv('../feature-importance-regression-cluster.csv')\n",
    "fts_importance_class = pd.read_csv('../feature-importance-regression-cluster-gpu.csv')\n",
    "fts_importance_reg_sig = fts_importance_reg[fts_importance_reg['normalized_importance'] > 0.01]\n",
    "fts_importance_class_sig = fts_importance_class[fts_importance_class['normalized_importance'] > 0.01]\n",
    "print(f'{len(fts_importance_reg_sig)} of {len(fts_importance_reg)} features are significant for regression.')\n",
    "print(f'{len(fts_importance_class_sig)} of {len(fts_importance_class)} features are significant for classification.')\n",
    "print(f'{len(set(fts_importance_class_sig[\"feature\"]).intersection(fts_importance_reg_sig[\"feature\"])) / len(fts_importance_reg):.2f} of features are significant for both regression and classification.')\n",
    "print(f'{len(set(fts_importance_class_sig[\"feature\"]).intersection(fts_importance_reg_sig[\"feature\"])) / len(set(fts_importance_class_sig[\"feature\"]).union(fts_importance_reg_sig[\"feature\"]))} of the significant features are significant for both regression and classification.')\n",
    "# (fts_importance_reg_sig.set_index('feature') - fts_importance_class_sig.set_index('feature')).reset_index()\n",
    "dfs = [fts_importance_class_sig, fts_importance_reg_sig]\n",
    "top_10_features = set().union(*[df[:10]['feature'].values for df in dfs])\n",
    "visualizations.slope_chart(dfs, labels=['Classification', 'Regression'], feature_selection=top_10_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib qt\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    sa = pd.read_csv('../results/spatial-autocorrelation-band-FRA.csv')\n",
    "    sns.lineplot(data=sa, x='distance', y='age', ci='sd', hue='country', ax=ax)\n",
    "    # ax.set(xscale='log')\n",
    "    ax.legend_.set_title(None)\n",
    "    ax.set_title('Spatial autocorrelation over distance')\n",
    "    ax.set_ylabel(\"Moran's I\")\n",
    "    ax.set_xlabel('distance [m]')\n",
    "    plt.savefig('../figures/spatial-autocorrelation-age.svg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-08-06 15:44:37,684 | WARNING : Dataset is already a GeoDataFrame. Skipping convertion.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['Staphorst', 'Bunschoten', 'Zijpe', 'Beuningen']\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb Cell 31'\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=8'>9</a>\u001b[0m aux_vars_geo\u001b[39m.\u001b[39mdropna(subset\u001b[39m=\u001b[39m[\u001b[39m'\u001b[39m\u001b[39mid\u001b[39m\u001b[39m'\u001b[39m, \u001b[39m'\u001b[39m\u001b[39mage\u001b[39m\u001b[39m'\u001b[39m], inplace\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m)\n\u001b[1;32m     <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=9'>10</a>\u001b[0m \u001b[39m# results = aux_vars_geo.groupby('city').apply(lambda x: spatial_autocorrelation.plot_correlogram_over_distance(x, attributes=['age', 'height', 'FootprintArea', 'Perimeter']))\u001b[39;00m\n\u001b[0;32m---> <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=10'>11</a>\u001b[0m results \u001b[39m=\u001b[39m aux_vars_geo\u001b[39m.\u001b[39;49mgroupby(\u001b[39m'\u001b[39;49m\u001b[39mcity\u001b[39;49m\u001b[39m'\u001b[39;49m)\u001b[39m.\u001b[39;49mapply(\u001b[39mlambda\u001b[39;49;00m x: spatial_autocorrelation\u001b[39m.\u001b[39;49mplot_correlogram_over_distance(x, attributes\u001b[39m=\u001b[39;49m[\u001b[39m'\u001b[39;49m\u001b[39mage\u001b[39;49m\u001b[39m'\u001b[39;49m], distances\u001b[39m=\u001b[39;49m[\u001b[39m10\u001b[39;49m, \u001b[39m25\u001b[39;49m, \u001b[39m50\u001b[39;49m, \u001b[39m100\u001b[39;49m, \u001b[39m250\u001b[39;49m, \u001b[39m500\u001b[39;49m, \u001b[39m750\u001b[39;49m, \u001b[39m1000\u001b[39;49m]))\n\u001b[1;32m     <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=11'>12</a>\u001b[0m results\u001b[39m.\u001b[39mto_csv(\u001b[39m'\u001b[39m\u001b[39m../results/spatial-autocorrelation-band.csv\u001b[39m\u001b[39m'\u001b[39m, index\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m)\n\u001b[1;32m     <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=12'>13</a>\u001b[0m results\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/groupby/groupby.py:1414\u001b[0m, in \u001b[0;36mGroupBy.apply\u001b[0;34m(self, func, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1412\u001b[0m \u001b[39mwith\u001b[39;00m option_context(\u001b[39m\"\u001b[39m\u001b[39mmode.chained_assignment\u001b[39m\u001b[39m\"\u001b[39m, \u001b[39mNone\u001b[39;00m):\n\u001b[1;32m   1413\u001b[0m     \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m-> 1414\u001b[0m         result \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_python_apply_general(f, \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_selected_obj)\n\u001b[1;32m   1415\u001b[0m     \u001b[39mexcept\u001b[39;00m \u001b[39mTypeError\u001b[39;00m:\n\u001b[1;32m   1416\u001b[0m         \u001b[39m# gh-20949\u001b[39;00m\n\u001b[1;32m   1417\u001b[0m         \u001b[39m# try again, with .apply acting as a filtering\u001b[39;00m\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   1421\u001b[0m         \u001b[39m# fails on *some* columns, e.g. a numeric operation\u001b[39;00m\n\u001b[1;32m   1422\u001b[0m         \u001b[39m# on a string grouper column\u001b[39;00m\n\u001b[1;32m   1424\u001b[0m         \u001b[39mwith\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_group_selection_context():\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/groupby/groupby.py:1455\u001b[0m, in \u001b[0;36mGroupBy._python_apply_general\u001b[0;34m(self, f, data, not_indexed_same)\u001b[0m\n\u001b[1;32m   1429\u001b[0m \u001b[39m@final\u001b[39m\n\u001b[1;32m   1430\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39m_python_apply_general\u001b[39m(\n\u001b[1;32m   1431\u001b[0m     \u001b[39mself\u001b[39m,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   1434\u001b[0m     not_indexed_same: \u001b[39mbool\u001b[39m \u001b[39m|\u001b[39m \u001b[39mNone\u001b[39;00m \u001b[39m=\u001b[39m \u001b[39mNone\u001b[39;00m,\n\u001b[1;32m   1435\u001b[0m ) \u001b[39m-\u001b[39m\u001b[39m>\u001b[39m DataFrame \u001b[39m|\u001b[39m Series:\n\u001b[1;32m   1436\u001b[0m     \u001b[39m\"\"\"\u001b[39;00m\n\u001b[1;32m   1437\u001b[0m \u001b[39m    Apply function f in python space\u001b[39;00m\n\u001b[1;32m   1438\u001b[0m \n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   1453\u001b[0m \u001b[39m        data after applying f\u001b[39;00m\n\u001b[1;32m   1454\u001b[0m \u001b[39m    \"\"\"\u001b[39;00m\n\u001b[0;32m-> 1455\u001b[0m     values, mutated \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mgrouper\u001b[39m.\u001b[39;49mapply(f, data, \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49maxis)\n\u001b[1;32m   1457\u001b[0m     \u001b[39mif\u001b[39;00m not_indexed_same \u001b[39mis\u001b[39;00m \u001b[39mNone\u001b[39;00m:\n\u001b[1;32m   1458\u001b[0m         not_indexed_same \u001b[39m=\u001b[39m mutated \u001b[39mor\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mmutated\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/groupby/ops.py:761\u001b[0m, in \u001b[0;36mBaseGrouper.apply\u001b[0;34m(self, f, data, axis)\u001b[0m\n\u001b[1;32m    759\u001b[0m \u001b[39m# group might be modified\u001b[39;00m\n\u001b[1;32m    760\u001b[0m group_axes \u001b[39m=\u001b[39m group\u001b[39m.\u001b[39maxes\n\u001b[0;32m--> 761\u001b[0m res \u001b[39m=\u001b[39m f(group)\n\u001b[1;32m    762\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mnot\u001b[39;00m mutated \u001b[39mand\u001b[39;00m \u001b[39mnot\u001b[39;00m _is_indexed_like(res, group_axes, axis):\n\u001b[1;32m    763\u001b[0m     mutated \u001b[39m=\u001b[39m \u001b[39mTrue\u001b[39;00m\n",
      "\u001b[1;32m/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb Cell 31'\u001b[0m in \u001b[0;36m<lambda>\u001b[0;34m(x)\u001b[0m\n\u001b[1;32m      <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=8'>9</a>\u001b[0m aux_vars_geo\u001b[39m.\u001b[39mdropna(subset\u001b[39m=\u001b[39m[\u001b[39m'\u001b[39m\u001b[39mid\u001b[39m\u001b[39m'\u001b[39m, \u001b[39m'\u001b[39m\u001b[39mage\u001b[39m\u001b[39m'\u001b[39m], inplace\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m)\n\u001b[1;32m     <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=9'>10</a>\u001b[0m \u001b[39m# results = aux_vars_geo.groupby('city').apply(lambda x: spatial_autocorrelation.plot_correlogram_over_distance(x, attributes=['age', 'height', 'FootprintArea', 'Perimeter']))\u001b[39;00m\n\u001b[0;32m---> <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=10'>11</a>\u001b[0m results \u001b[39m=\u001b[39m aux_vars_geo\u001b[39m.\u001b[39mgroupby(\u001b[39m'\u001b[39m\u001b[39mcity\u001b[39m\u001b[39m'\u001b[39m)\u001b[39m.\u001b[39mapply(\u001b[39mlambda\u001b[39;00m x: spatial_autocorrelation\u001b[39m.\u001b[39;49mplot_correlogram_over_distance(x, attributes\u001b[39m=\u001b[39;49m[\u001b[39m'\u001b[39;49m\u001b[39mage\u001b[39;49m\u001b[39m'\u001b[39;49m], distances\u001b[39m=\u001b[39;49m[\u001b[39m10\u001b[39;49m, \u001b[39m25\u001b[39;49m, \u001b[39m50\u001b[39;49m, \u001b[39m100\u001b[39;49m, \u001b[39m250\u001b[39;49m, \u001b[39m500\u001b[39;49m, \u001b[39m750\u001b[39;49m, \u001b[39m1000\u001b[39;49m]))\n\u001b[1;32m     <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=11'>12</a>\u001b[0m results\u001b[39m.\u001b[39mto_csv(\u001b[39m'\u001b[39m\u001b[39m../results/spatial-autocorrelation-band.csv\u001b[39m\u001b[39m'\u001b[39m, index\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m)\n\u001b[1;32m     <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000332?line=12'>13</a>\u001b[0m results\n",
      "File \u001b[0;32m~/workspace/ufo-prediction/notebooks/../src/spatial_autocorrelation.py:108\u001b[0m, in \u001b[0;36mplot_correlogram_over_distance\u001b[0;34m(df, attributes, distances)\u001b[0m\n\u001b[1;32m    105\u001b[0m neighbors \u001b[39m=\u001b[39m {}\n\u001b[1;32m    106\u001b[0m \u001b[39mfor\u001b[39;00m dis \u001b[39min\u001b[39;00m distances:\n\u001b[1;32m    107\u001b[0m     \u001b[39m# weights = _distance_weights(df, distance_threshold=dis)\u001b[39;00m\n\u001b[0;32m--> 108\u001b[0m     weights \u001b[39m=\u001b[39m _distance_band_weights(df, distance_threshold\u001b[39m=\u001b[39;49mdis, neighbors_to_exclude\u001b[39m=\u001b[39;49mneighbors)\n\u001b[1;32m    109\u001b[0m     neighbors \u001b[39m=\u001b[39m weights\u001b[39m.\u001b[39mneighbors\n\u001b[1;32m    110\u001b[0m     coefficients\u001b[39m.\u001b[39mappend({attr: Moran(df[attr], weights)\u001b[39m.\u001b[39mI \u001b[39mfor\u001b[39;00m attr \u001b[39min\u001b[39;00m attributes})\n",
      "File \u001b[0;32m~/workspace/ufo-prediction/notebooks/../src/spatial_autocorrelation.py:161\u001b[0m, in \u001b[0;36m_distance_band_weights\u001b[0;34m(gdf, distance_threshold, neighbors_to_exclude)\u001b[0m\n\u001b[1;32m    160\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39m_distance_band_weights\u001b[39m(gdf, distance_threshold, neighbors_to_exclude):\n\u001b[0;32m--> 161\u001b[0m     dis \u001b[39m=\u001b[39m lps\u001b[39m.\u001b[39;49mweights\u001b[39m.\u001b[39;49mDistanceBand\u001b[39m.\u001b[39;49mfrom_dataframe(gdf, threshold\u001b[39m=\u001b[39;49mdistance_threshold, ids\u001b[39m=\u001b[39;49mgdf[\u001b[39m'\u001b[39;49m\u001b[39mid\u001b[39;49m\u001b[39m'\u001b[39;49m]\u001b[39m.\u001b[39;49mvalues, silence_warnings\u001b[39m=\u001b[39;49m\u001b[39mTrue\u001b[39;49;00m)\n\u001b[1;32m    162\u001b[0m     neighbors \u001b[39m=\u001b[39m {}\n\u001b[1;32m    163\u001b[0m     \u001b[39mfor\u001b[39;00m building_id, neighbor_ids \u001b[39min\u001b[39;00m dis\u001b[39m.\u001b[39mneighbors\u001b[39m.\u001b[39mitems():\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/libpysal/weights/distance.py:910\u001b[0m, in \u001b[0;36mDistanceBand.from_dataframe\u001b[0;34m(cls, df, threshold, geom_col, ids, **kwargs)\u001b[0m\n\u001b[1;32m    908\u001b[0m \u001b[39melif\u001b[39;00m \u001b[39misinstance\u001b[39m(ids, \u001b[39mstr\u001b[39m):\n\u001b[1;32m    909\u001b[0m     ids \u001b[39m=\u001b[39m df[ids]\u001b[39m.\u001b[39mtolist()\n\u001b[0;32m--> 910\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mcls\u001b[39;49m(pts, threshold, ids\u001b[39m=\u001b[39;49mids, \u001b[39m*\u001b[39;49m\u001b[39m*\u001b[39;49mkwargs)\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/libpysal/weights/distance.py:844\u001b[0m, in \u001b[0;36mDistanceBand.__init__\u001b[0;34m(self, data, threshold, p, alpha, binary, ids, build_sp, silence_warnings, distance_metric, radius)\u001b[0m\n\u001b[1;32m    841\u001b[0m         \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mdata \u001b[39m=\u001b[39m data\n\u001b[1;32m    842\u001b[0m         \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mkdtree \u001b[39m=\u001b[39m \u001b[39mNone\u001b[39;00m\n\u001b[0;32m--> 844\u001b[0m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_band()\n\u001b[1;32m    845\u001b[0m neighbors, weights \u001b[39m=\u001b[39m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_distance_to_W(ids)\n\u001b[1;32m    846\u001b[0m W\u001b[39m.\u001b[39m\u001b[39m__init__\u001b[39m(\n\u001b[1;32m    847\u001b[0m     \u001b[39mself\u001b[39m, neighbors, weights, ids, silence_warnings\u001b[39m=\u001b[39m\u001b[39mself\u001b[39m\u001b[39m.\u001b[39msilence_warnings\n\u001b[1;32m    848\u001b[0m )\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/libpysal/weights/distance.py:915\u001b[0m, in \u001b[0;36mDistanceBand._band\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    913\u001b[0m \u001b[39m\"\"\"Find all pairs within threshold.\"\"\"\u001b[39;00m\n\u001b[1;32m    914\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mbuild_sp:\n\u001b[0;32m--> 915\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mdmat \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mkdtree\u001b[39m.\u001b[39;49msparse_distance_matrix(\n\u001b[1;32m    916\u001b[0m         \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mkdtree, max_distance\u001b[39m=\u001b[39;49m\u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mthreshold, p\u001b[39m=\u001b[39;49m\u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mp\n\u001b[1;32m    917\u001b[0m     )\u001b[39m.\u001b[39mtocsr()\n\u001b[1;32m    918\u001b[0m \u001b[39melse\u001b[39;00m:\n\u001b[1;32m    919\u001b[0m     \u001b[39mif\u001b[39;00m \u001b[39mstr\u001b[39m(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mkdtree)\u001b[39m.\u001b[39msplit(\u001b[39m\"\u001b[39m\u001b[39m.\u001b[39m\u001b[39m\"\u001b[39m)[\u001b[39m-\u001b[39m\u001b[39m1\u001b[39m][\u001b[39m0\u001b[39m:\u001b[39m10\u001b[39m] \u001b[39m==\u001b[39m \u001b[39m\"\u001b[39m\u001b[39mArc_KDTree\u001b[39m\u001b[39m\"\u001b[39m:\n",
      "File \u001b[0;32m_ckdtree.pyx:1545\u001b[0m, in \u001b[0;36mscipy.spatial._ckdtree.cKDTree.sparse_distance_matrix\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32m_ckdtree.pyx:235\u001b[0m, in \u001b[0;36mscipy.spatial._ckdtree.coo_entries.dok_matrix\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/scipy/sparse/_coo.py:442\u001b[0m, in \u001b[0;36mcoo_matrix.todok\u001b[0;34m(self, copy)\u001b[0m\n\u001b[1;32m    441\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mtodok\u001b[39m(\u001b[39mself\u001b[39m, copy\u001b[39m=\u001b[39m\u001b[39mFalse\u001b[39;00m):\n\u001b[0;32m--> 442\u001b[0m     \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49msum_duplicates()\n\u001b[1;32m    443\u001b[0m     dok \u001b[39m=\u001b[39m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_dok_container((\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mshape), dtype\u001b[39m=\u001b[39m\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mdtype)\n\u001b[1;32m    444\u001b[0m     dok\u001b[39m.\u001b[39m_update(\u001b[39mzip\u001b[39m(\u001b[39mzip\u001b[39m(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mrow,\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mcol),\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mdata))\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/scipy/sparse/_coo.py:527\u001b[0m, in \u001b[0;36mcoo_matrix.sum_duplicates\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    525\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mhas_canonical_format:\n\u001b[1;32m    526\u001b[0m     \u001b[39mreturn\u001b[39;00m\n\u001b[0;32m--> 527\u001b[0m summed \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_sum_duplicates(\u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mrow, \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mcol, \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mdata)\n\u001b[1;32m    528\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mrow, \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mcol, \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mdata \u001b[39m=\u001b[39m summed\n\u001b[1;32m    529\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mhas_canonical_format \u001b[39m=\u001b[39m \u001b[39mTrue\u001b[39;00m\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/scipy/sparse/_coo.py:535\u001b[0m, in \u001b[0;36mcoo_matrix._sum_duplicates\u001b[0;34m(self, row, col, data)\u001b[0m\n\u001b[1;32m    533\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mlen\u001b[39m(data) \u001b[39m==\u001b[39m \u001b[39m0\u001b[39m:\n\u001b[1;32m    534\u001b[0m     \u001b[39mreturn\u001b[39;00m row, col, data\n\u001b[0;32m--> 535\u001b[0m order \u001b[39m=\u001b[39m np\u001b[39m.\u001b[39;49mlexsort((row, col))\n\u001b[1;32m    536\u001b[0m row \u001b[39m=\u001b[39m row[order]\n\u001b[1;32m    537\u001b[0m col \u001b[39m=\u001b[39m col[order]\n",
      "File \u001b[0;32m<__array_function__ internals>:5\u001b[0m, in \u001b[0;36mlexsort\u001b[0;34m(*args, **kwargs)\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "\n",
    "# for country, df_path in {'ESP': ESP_DATA_PATH}.items():\n",
    "df = df_nl_pre\n",
    "cities = df['city'].value_counts().between(5000, 10000)\n",
    "cities = list(cities[cities].index)[:4]\n",
    "print(cities)\n",
    "df = df[df['city'].isin(cities)]\n",
    "aux_vars_geo = geometry.to_gdf(df[['id', 'geometry', 'neighborhood', 'sbb', 'city', 'country', 'height', 'age', 'FootprintArea', 'Perimeter']])\n",
    "# aux_vars_geo.dropna(subset=['id', 'age', 'height', 'FootprintArea', 'Perimeter'], inplace=True)\n",
    "aux_vars_geo.dropna(subset=['id', 'age'], inplace=True)\n",
    "# results = aux_vars_geo.groupby('city').apply(lambda x: spatial_autocorrelation.plot_correlogram_over_distance(x, attributes=['age', 'height', 'FootprintArea', 'Perimeter']))\n",
    "results = aux_vars_geo.groupby('city').apply(lambda x: spatial_autocorrelation.plot_correlogram_over_distance(x, attributes=['age'], distances=[10, 25, 50, 100, 250, 500, 750, 1000]))\n",
    "results.to_csv('../results/spatial-autocorrelation-band.csv', index=True)\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Amsterdam              131694\n",
       "Apeldoorn              102134\n",
       "Woerden                 30175\n",
       "Dronten                 22086\n",
       "Alphen aan den Rijn     20504\n",
       "De Ronde Venen          20314\n",
       "Veere                   19276\n",
       "Borger-Odoorn           19200\n",
       "Voorst                  18996\n",
       "Deurne                  18236\n",
       "Tytsjerksteradiel       18180\n",
       "Dinkelland              18168\n",
       "Weststellingwerf        17436\n",
       "IJsselstein             15495\n",
       "Vlissingen              15185\n",
       "Brunssum                14132\n",
       "Geertruidenberg         12215\n",
       "Bloemendaal             11526\n",
       "Sint-Oedenrode          11262\n",
       "Binnenmaas              11236\n",
       "Buren                   10938\n",
       "Groesbeek               10643\n",
       "Goedereede              10413\n",
       "Staphorst                9458\n",
       "Bunschoten               8915\n",
       "Zijpe                    7508\n",
       "Beuningen                7258\n",
       "Brielle                  7011\n",
       "Het Bildt                6573\n",
       "Drimmelen                6505\n",
       "Rijnwaarden              6330\n",
       "Veenendaal               5999\n",
       "Roggel en Neer           5542\n",
       "Blaricum                 5443\n",
       "Maasdonk                 4574\n",
       "Renkum                   4437\n",
       "Valkenswaard             4193\n",
       "Oldenzaal                3951\n",
       "Uithoorn                 3820\n",
       "Dinxperlo                3610\n",
       "Jacobswoude              3593\n",
       "Kapelle                  2449\n",
       "Nuth                     2439\n",
       "Vianen                   2368\n",
       "Drechterland              351\n",
       "Bathmen                   210\n",
       "Name: city, dtype: int64"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cities = df_nl_pre['city'].value_counts().between(100, 1000)\n",
    "\n",
    "cities[cities].index\n",
    "df_nl_pre['city'].value_counts()\n",
    "# a[a < 1000]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-08-06 11:32:52,405 | WARNING : Dataset is already a GeoDataFrame. Skipping convertion.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>height</th>\n",
       "      <th>FootprintArea</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>city</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">Bathmen</th>\n",
       "      <th>10</th>\n",
       "      <td>0.487978</td>\n",
       "      <td>1.186010</td>\n",
       "      <td>0.330211</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.562608</td>\n",
       "      <td>0.100382</td>\n",
       "      <td>0.124832</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">Drechterland</th>\n",
       "      <th>10</th>\n",
       "      <td>0.466180</td>\n",
       "      <td>0.369090</td>\n",
       "      <td>0.117502</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.458970</td>\n",
       "      <td>0.090431</td>\n",
       "      <td>0.064787</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                      age    height  FootprintArea\n",
       "city                                              \n",
       "Bathmen      10  0.487978  1.186010       0.330211\n",
       "             25  0.562608  0.100382       0.124832\n",
       "Drechterland 10  0.466180  0.369090       0.117502\n",
       "             25  0.458970  0.090431       0.064787"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "aux_vars_geo = geometry.to_gdf(df_nl_pre[['geometry', 'neighborhood', 'sbb', 'city', 'country', 'height', 'age', 'FootprintArea']])\n",
    "aux_vars_geo.dropna(subset=['age', 'height', 'FootprintArea'], inplace=True)\n",
    "aux_vars_geo = aux_vars_geo[aux_vars_geo['city'].isin(['Drechterland', 'Bathmen'])]\n",
    "aux_vars_geo.groupby('city').apply(lambda x: spatial_autocorrelation.plot_correlogram_over_distance(x, attributes=['age', 'height', 'FootprintArea'], distances=[10, 25]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-08-05 20:48:36,082 | WARNING : Dataset is already a GeoDataFrame. Skipping convertion.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>height</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.590248</td>\n",
       "      <td>0.317733</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.506469</td>\n",
       "      <td>0.439570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>0.457080</td>\n",
       "      <td>0.475788</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         age    height\n",
       "10  0.590248  0.317733\n",
       "25  0.506469  0.439570\n",
       "50  0.457080  0.475788"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictor = regressor_fts_importance_fr\n",
    "# y = predictor.y_test.rename(columns={'age': 'y_test'})\n",
    "aux_vars = pd.concat([predictor.aux_vars_test, predictor.y_test], axis=1, join=\"inner\")\n",
    "aux_vars = aux_vars[aux_vars['city'] == 'Toulouse']\n",
    "\n",
    "# aux_vars_geo = geometry.lat_lon_to_gdf(aux_vars[['lat', 'lon', 'neighborhood', 'sbb', 'city', 'country', 'height', 'age']])\n",
    "aux_vars_geo = geometry.to_gdf(aux_vars[['geometry', 'neighborhood', 'sbb', 'city', 'country', 'height', 'age']])\n",
    "aux_vars_geo.dropna(subset=['age', 'height'], inplace=True)\n",
    "aux_vars_geo\n",
    "spatial_autocorrelation.plot_correlogram_over_distance(aux_vars_geo, attributes=['age', 'height'], distances=[10, 25, 50])\n",
    "# spatial_autocorrelation.plot_correlogram_over_distance()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Hyperparameter tuning - local test run"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-06-28 15:58:45,725 | INFO : Generated bins: [0, 1915, 1949, 1968, 1975, 1982, 1990, 2000, 2006, 2013, 2051]\n",
      "2022-06-28 15:58:45,726 | INFO : Generated bins with the following labels: ['<1915', '1915-1948', '1949-1967', '1968-1974', '1975-1981', '1982-1989', '1990-1999', '2000-2005', '2006-2012', '2013-2050']\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-06-28 15:58:45,763 | INFO : Dataset length: 4441\n",
      "2022-06-28 15:58:45,769 | INFO : Dataset allocated memory: 4 MB\n",
      "2022-06-28 15:58:45,776 | INFO : Training dataset length: 2960\n",
      "2022-06-28 15:58:45,776 | INFO : Test dataset length: 1481\n",
      "2022-06-28 15:58:45,789 | INFO : age attribute has been categorized (lowest age included: 0; highest age included: 2050).\n",
      "2022-06-28 15:58:45,796 | INFO : age attribute has been categorized (lowest age included: 0; highest age included: 2050).\n",
      "2022-06-28 15:58:45,811 | INFO : Training dataset length after preprocessing: 2070\n",
      "2022-06-28 15:58:45,812 | INFO : Test dataset length after preprocessing: 1004\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-06-28 15:58:50,225 | INFO : Training dataset length: 2961\n",
      "2022-06-28 15:58:50,226 | INFO : Test dataset length: 1480\n",
      "2022-06-28 15:58:50,238 | INFO : age attribute has been categorized (lowest age included: 0; highest age included: 2050).\n",
      "2022-06-28 15:58:50,244 | INFO : age attribute has been categorized (lowest age included: 0; highest age included: 2050).\n",
      "2022-06-28 15:58:50,256 | INFO : Training dataset length after preprocessing: 2035\n",
      "2022-06-28 15:58:50,257 | INFO : Test dataset length after preprocessing: 1039\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-06-28 15:58:54,369 | INFO : Training dataset length: 2961\n",
      "2022-06-28 15:58:54,370 | INFO : Test dataset length: 1480\n",
      "2022-06-28 15:58:54,381 | INFO : age attribute has been categorized (lowest age included: 0; highest age included: 2050).\n",
      "2022-06-28 15:58:54,389 | INFO : age attribute has been categorized (lowest age included: 0; highest age included: 2050).\n",
      "2022-06-28 15:58:54,406 | INFO : Training dataset length after preprocessing: 2043\n",
      "2022-06-28 15:58:54,407 | INFO : Test dataset length after preprocessing: 1031\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-06-28 15:58:59,161 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 15:58:59,190 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 15:58:59,203 | INFO : R2: 0.3780\n",
      "2022-06-28 15:58:59,204 | INFO : MAPE: 0.4327\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification report:\n",
      "               precision    recall  f1-score  support\n",
      "<1915          0.333333  0.311765  0.322188      170\n",
      "1915-1948      0.549533  0.597561  0.572541      492\n",
      "1949-1967      0.507853  0.565049  0.534926      515\n",
      "1968-1974      0.309783  0.234568  0.266979      243\n",
      "1975-1981      0.513619  0.534413  0.523810      494\n",
      "1982-1989      0.524079  0.506849  0.515320      365\n",
      "1990-1999      0.488055  0.451104  0.468852      317\n",
      "2000-2005      0.445455  0.471154  0.457944      208\n",
      "2006-2012      0.357895  0.363636  0.360743      187\n",
      "2013-2050      0.584906  0.373494  0.455882       83\n",
      "accuracy       0.482759  0.482759  0.482759        0\n",
      "macro avg      0.461451  0.440959  0.447919     3074\n",
      "weighted avg   0.478763  0.482759  0.479133     3074\n",
      "Cohen’s kappa: 0.4081\n",
      "Matthews correlation coefficient (MCC): 0.4084\n",
      "R2: 0.3780\n",
      "MAPE: 0.4327\n"
     ]
    }
   ],
   "source": [
    "df_fr_sample['country'] = 'France'\n",
    "df_nl_sample['country'] = 'Netherlands'\n",
    "predictor = AgeClassifier(\n",
    "    model=XGBClassifier(tree_method='hist', n_jobs=-1),\n",
    "    df=df_fr_sample,\n",
    "    # test_set=df_fr_sample,\n",
    "    # test_training_split=split_80_20,\n",
    "    cross_validation_split=cross_validation,\n",
    "    # hyperparameter_tuning_only=True,\n",
    "    # hyperparameter_tuning=True,\n",
    "    early_stopping=True,\n",
    "    # hyperparameters=hyperparameters,\n",
    "    preprocessing_stages=[remove_outliers],\n",
    "    bins=dataset.TABULA_AGE_BINS['france'],\n",
    "    mitigate_class_imbalance=True,\n",
    ")\n",
    "predictor.print_model_error()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-06-28 15:59:53,147 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 15:59:53,171 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 15:59:53,189 | INFO : R2: 0.3780\n",
      "2022-06-28 15:59:53,201 | INFO : MAPE: 0.4327\n",
      "2022-06-28 15:59:53,230 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 15:59:53,236 | ERROR : Failed to calculate energy error: cannot set a frame with no defined index and a scalar\n",
      "2022-06-28 15:59:53,285 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 15:59:53,314 | ERROR : Failed to calculate energy error: cannot set a frame with no defined index and a scalar\n",
      "2022-06-28 15:59:53,354 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 15:59:53,385 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 15:59:53,397 | INFO : R2: 0.3780\n",
      "2022-06-28 15:59:53,398 | INFO : MAPE: 0.4327\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "      <th>support</th>\n",
       "      <th>kappa</th>\n",
       "      <th>mcc</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>energy_r2</th>\n",
       "      <th>energy_mape</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>&lt;1915</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.311765</td>\n",
       "      <td>0.322188</td>\n",
       "      <td>170.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1915-1948</th>\n",
       "      <td>0.549533</td>\n",
       "      <td>0.597561</td>\n",
       "      <td>0.572541</td>\n",
       "      <td>492.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1949-1967</th>\n",
       "      <td>0.507853</td>\n",
       "      <td>0.565049</td>\n",
       "      <td>0.534926</td>\n",
       "      <td>515.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1968-1974</th>\n",
       "      <td>0.309783</td>\n",
       "      <td>0.234568</td>\n",
       "      <td>0.266979</td>\n",
       "      <td>243.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1975-1981</th>\n",
       "      <td>0.513619</td>\n",
       "      <td>0.534413</td>\n",
       "      <td>0.523810</td>\n",
       "      <td>494.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1982-1989</th>\n",
       "      <td>0.524079</td>\n",
       "      <td>0.506849</td>\n",
       "      <td>0.515320</td>\n",
       "      <td>365.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1990-1999</th>\n",
       "      <td>0.488055</td>\n",
       "      <td>0.451104</td>\n",
       "      <td>0.468852</td>\n",
       "      <td>317.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2000-2005</th>\n",
       "      <td>0.445455</td>\n",
       "      <td>0.471154</td>\n",
       "      <td>0.457944</td>\n",
       "      <td>208.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2006-2012</th>\n",
       "      <td>0.357895</td>\n",
       "      <td>0.363636</td>\n",
       "      <td>0.360743</td>\n",
       "      <td>187.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2013-2050</th>\n",
       "      <td>0.584906</td>\n",
       "      <td>0.373494</td>\n",
       "      <td>0.455882</td>\n",
       "      <td>83.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>macro avg</th>\n",
       "      <td>0.461451</td>\n",
       "      <td>0.440959</td>\n",
       "      <td>0.447919</td>\n",
       "      <td>3074.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>weighted avg</th>\n",
       "      <td>0.478763</td>\n",
       "      <td>0.482759</td>\n",
       "      <td>0.479133</td>\n",
       "      <td>3074.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>total</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.408137</td>\n",
       "      <td>0.408413</td>\n",
       "      <td>0.482759</td>\n",
       "      <td>0.377989</td>\n",
       "      <td>0.432668</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fold_0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.395510</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fold_1</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.424763</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fold_2</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.405340</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.377989</td>\n",
       "      <td>0.432668</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              precision    recall  f1-score  support     kappa       mcc  \\\n",
       "<1915          0.333333  0.311765  0.322188    170.0       NaN       NaN   \n",
       "1915-1948      0.549533  0.597561  0.572541    492.0       NaN       NaN   \n",
       "1949-1967      0.507853  0.565049  0.534926    515.0       NaN       NaN   \n",
       "1968-1974      0.309783  0.234568  0.266979    243.0       NaN       NaN   \n",
       "1975-1981      0.513619  0.534413  0.523810    494.0       NaN       NaN   \n",
       "1982-1989      0.524079  0.506849  0.515320    365.0       NaN       NaN   \n",
       "1990-1999      0.488055  0.451104  0.468852    317.0       NaN       NaN   \n",
       "2000-2005      0.445455  0.471154  0.457944    208.0       NaN       NaN   \n",
       "2006-2012      0.357895  0.363636  0.360743    187.0       NaN       NaN   \n",
       "2013-2050      0.584906  0.373494  0.455882     83.0       NaN       NaN   \n",
       "macro avg      0.461451  0.440959  0.447919   3074.0       NaN       NaN   \n",
       "weighted avg   0.478763  0.482759  0.479133   3074.0       NaN       NaN   \n",
       "total               NaN       NaN       NaN      NaN  0.408137  0.408413   \n",
       "fold_0              NaN       NaN       NaN      NaN       NaN  0.395510   \n",
       "fold_1              NaN       NaN       NaN      NaN       NaN  0.424763   \n",
       "fold_2              NaN       NaN       NaN      NaN       NaN  0.405340   \n",
       "\n",
       "              accuracy  energy_r2  energy_mape  \n",
       "<1915              NaN        NaN          NaN  \n",
       "1915-1948          NaN        NaN          NaN  \n",
       "1949-1967          NaN        NaN          NaN  \n",
       "1968-1974          NaN        NaN          NaN  \n",
       "1975-1981          NaN        NaN          NaN  \n",
       "1982-1989          NaN        NaN          NaN  \n",
       "1990-1999          NaN        NaN          NaN  \n",
       "2000-2005          NaN        NaN          NaN  \n",
       "2006-2012          NaN        NaN          NaN  \n",
       "2013-2050          NaN        NaN          NaN  \n",
       "macro avg          NaN        NaN          NaN  \n",
       "weighted avg       NaN        NaN          NaN  \n",
       "total         0.482759   0.377989     0.432668  \n",
       "fold_0             NaN        NaN          NaN  \n",
       "fold_1             NaN        NaN          NaN  \n",
       "fold_2             NaN   0.377989     0.432668  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictor.error()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-06-28 16:18:40,823 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 16:18:40,844 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 16:18:40,854 | INFO : R2: 0.3780\n",
      "2022-06-28 16:18:40,855 | INFO : MAPE: 0.4327\n",
      "2022-06-28 16:18:40,868 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 16:18:40,869 | ERROR : Failed to calculate energy error: cannot set a frame with no defined index and a scalar\n",
      "2022-06-28 16:18:40,881 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 16:18:40,882 | ERROR : Failed to calculate energy error: cannot set a frame with no defined index and a scalar\n",
      "2022-06-28 16:18:40,907 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 16:18:40,934 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-06-28 16:18:40,946 | INFO : R2: 0.3780\n",
      "2022-06-28 16:18:40,946 | INFO : MAPE: 0.4327\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "      <th>support</th>\n",
       "      <th>kappa</th>\n",
       "      <th>mcc</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>energy_r2</th>\n",
       "      <th>energy_mape</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>&lt;1915</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.311765</td>\n",
       "      <td>0.322188</td>\n",
       "      <td>170.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1915-1948</th>\n",
       "      <td>0.549533</td>\n",
       "      <td>0.597561</td>\n",
       "      <td>0.572541</td>\n",
       "      <td>492.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1949-1967</th>\n",
       "      <td>0.507853</td>\n",
       "      <td>0.565049</td>\n",
       "      <td>0.534926</td>\n",
       "      <td>515.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1968-1974</th>\n",
       "      <td>0.309783</td>\n",
       "      <td>0.234568</td>\n",
       "      <td>0.266979</td>\n",
       "      <td>243.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1975-1981</th>\n",
       "      <td>0.513619</td>\n",
       "      <td>0.534413</td>\n",
       "      <td>0.523810</td>\n",
       "      <td>494.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1982-1989</th>\n",
       "      <td>0.524079</td>\n",
       "      <td>0.506849</td>\n",
       "      <td>0.515320</td>\n",
       "      <td>365.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1990-1999</th>\n",
       "      <td>0.488055</td>\n",
       "      <td>0.451104</td>\n",
       "      <td>0.468852</td>\n",
       "      <td>317.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2000-2005</th>\n",
       "      <td>0.445455</td>\n",
       "      <td>0.471154</td>\n",
       "      <td>0.457944</td>\n",
       "      <td>208.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2006-2012</th>\n",
       "      <td>0.357895</td>\n",
       "      <td>0.363636</td>\n",
       "      <td>0.360743</td>\n",
       "      <td>187.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2013-2050</th>\n",
       "      <td>0.584906</td>\n",
       "      <td>0.373494</td>\n",
       "      <td>0.455882</td>\n",
       "      <td>83.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>macro avg</th>\n",
       "      <td>0.461451</td>\n",
       "      <td>0.440959</td>\n",
       "      <td>0.447919</td>\n",
       "      <td>3074.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>weighted avg</th>\n",
       "      <td>0.478763</td>\n",
       "      <td>0.482759</td>\n",
       "      <td>0.479133</td>\n",
       "      <td>3074.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>total</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.408137</td>\n",
       "      <td>0.408413</td>\n",
       "      <td>0.482759</td>\n",
       "      <td>0.377989</td>\n",
       "      <td>0.432668</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fold_0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.395510</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fold_1</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.424763</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fold_2</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.405340</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.377989</td>\n",
       "      <td>0.432668</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              precision    recall  f1-score  support     kappa       mcc  \\\n",
       "<1915          0.333333  0.311765  0.322188    170.0       NaN       NaN   \n",
       "1915-1948      0.549533  0.597561  0.572541    492.0       NaN       NaN   \n",
       "1949-1967      0.507853  0.565049  0.534926    515.0       NaN       NaN   \n",
       "1968-1974      0.309783  0.234568  0.266979    243.0       NaN       NaN   \n",
       "1975-1981      0.513619  0.534413  0.523810    494.0       NaN       NaN   \n",
       "1982-1989      0.524079  0.506849  0.515320    365.0       NaN       NaN   \n",
       "1990-1999      0.488055  0.451104  0.468852    317.0       NaN       NaN   \n",
       "2000-2005      0.445455  0.471154  0.457944    208.0       NaN       NaN   \n",
       "2006-2012      0.357895  0.363636  0.360743    187.0       NaN       NaN   \n",
       "2013-2050      0.584906  0.373494  0.455882     83.0       NaN       NaN   \n",
       "macro avg      0.461451  0.440959  0.447919   3074.0       NaN       NaN   \n",
       "weighted avg   0.478763  0.482759  0.479133   3074.0       NaN       NaN   \n",
       "total               NaN       NaN       NaN      NaN  0.408137  0.408413   \n",
       "fold_0              NaN       NaN       NaN      NaN       NaN  0.395510   \n",
       "fold_1              NaN       NaN       NaN      NaN       NaN  0.424763   \n",
       "fold_2              NaN       NaN       NaN      NaN       NaN  0.405340   \n",
       "\n",
       "              accuracy  energy_r2  energy_mape  \n",
       "<1915              NaN        NaN          NaN  \n",
       "1915-1948          NaN        NaN          NaN  \n",
       "1949-1967          NaN        NaN          NaN  \n",
       "1968-1974          NaN        NaN          NaN  \n",
       "1975-1981          NaN        NaN          NaN  \n",
       "1982-1989          NaN        NaN          NaN  \n",
       "1990-1999          NaN        NaN          NaN  \n",
       "2000-2005          NaN        NaN          NaN  \n",
       "2006-2012          NaN        NaN          NaN  \n",
       "2013-2050          NaN        NaN          NaN  \n",
       "macro avg          NaN        NaN          NaN  \n",
       "weighted avg       NaN        NaN          NaN  \n",
       "total         0.482759   0.377989     0.432668  \n",
       "fold_0             NaN        NaN          NaN  \n",
       "fold_1             NaN        NaN          NaN  \n",
       "fold_2             NaN   0.377989     0.432668  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# predictor.rmse(across_folds=True)\n",
    "predictor.eval_metrics()\n",
    "# predictor.rmse()\n",
    "# residuals = predictor.individual_prediction_error()['error'].abs()\n",
    "# np.histogram(residuals, bins=[0, 5, 10, 20, np.inf])[0] / len(residuals)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Hyperparameter tuning - large-scale results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 624,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mean_test_score</th>\n",
       "      <th>std_test_score</th>\n",
       "      <th>mean_train_score</th>\n",
       "      <th>mean_fit_time</th>\n",
       "      <th>param_n_estimators</th>\n",
       "      <th>param_max_depth</th>\n",
       "      <th>param_learning_rate</th>\n",
       "      <th>param_colsample_bytree</th>\n",
       "      <th>param_colsample_bylevel</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.815052</td>\n",
       "      <td>0.010096</td>\n",
       "      <td>0.999442</td>\n",
       "      <td>44.528760</td>\n",
       "      <td>800</td>\n",
       "      <td>12</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.426461</td>\n",
       "      <td>0.006985</td>\n",
       "      <td>0.845973</td>\n",
       "      <td>3.892007</td>\n",
       "      <td>500</td>\n",
       "      <td>7</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>79</th>\n",
       "      <td>0.453990</td>\n",
       "      <td>0.010732</td>\n",
       "      <td>0.970382</td>\n",
       "      <td>34.947673</td>\n",
       "      <td>900</td>\n",
       "      <td>10</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>57</th>\n",
       "      <td>0.453582</td>\n",
       "      <td>0.010630</td>\n",
       "      <td>0.976138</td>\n",
       "      <td>41.552435</td>\n",
       "      <td>700</td>\n",
       "      <td>11</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>120</th>\n",
       "      <td>0.812194</td>\n",
       "      <td>0.013511</td>\n",
       "      <td>0.999468</td>\n",
       "      <td>45.108047</td>\n",
       "      <td>800</td>\n",
       "      <td>12</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>117</th>\n",
       "      <td>0.818274</td>\n",
       "      <td>0.007815</td>\n",
       "      <td>0.997005</td>\n",
       "      <td>27.323182</td>\n",
       "      <td>700</td>\n",
       "      <td>11</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>84</th>\n",
       "      <td>0.432648</td>\n",
       "      <td>0.016297</td>\n",
       "      <td>0.942675</td>\n",
       "      <td>13.140939</td>\n",
       "      <td>1000</td>\n",
       "      <td>8</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>100</th>\n",
       "      <td>0.817208</td>\n",
       "      <td>0.007655</td>\n",
       "      <td>0.999240</td>\n",
       "      <td>47.867464</td>\n",
       "      <td>800</td>\n",
       "      <td>12</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>60</th>\n",
       "      <td>0.453670</td>\n",
       "      <td>0.011622</td>\n",
       "      <td>0.993961</td>\n",
       "      <td>71.353427</td>\n",
       "      <td>800</td>\n",
       "      <td>12</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.421583</td>\n",
       "      <td>0.005526</td>\n",
       "      <td>0.999723</td>\n",
       "      <td>63.789768</td>\n",
       "      <td>700</td>\n",
       "      <td>13</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>139</th>\n",
       "      <td>0.811933</td>\n",
       "      <td>0.013830</td>\n",
       "      <td>0.997609</td>\n",
       "      <td>24.026194</td>\n",
       "      <td>900</td>\n",
       "      <td>10</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>0.814785</td>\n",
       "      <td>0.009315</td>\n",
       "      <td>0.997608</td>\n",
       "      <td>23.571766</td>\n",
       "      <td>900</td>\n",
       "      <td>10</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>90</th>\n",
       "      <td>0.432581</td>\n",
       "      <td>0.010002</td>\n",
       "      <td>0.825493</td>\n",
       "      <td>5.935951</td>\n",
       "      <td>800</td>\n",
       "      <td>7</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59</th>\n",
       "      <td>0.453180</td>\n",
       "      <td>0.010499</td>\n",
       "      <td>0.970333</td>\n",
       "      <td>34.490594</td>\n",
       "      <td>900</td>\n",
       "      <td>10</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.452273</td>\n",
       "      <td>0.010612</td>\n",
       "      <td>0.994166</td>\n",
       "      <td>71.622012</td>\n",
       "      <td>800</td>\n",
       "      <td>12</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>121</th>\n",
       "      <td>0.811535</td>\n",
       "      <td>0.014641</td>\n",
       "      <td>0.999844</td>\n",
       "      <td>34.210339</td>\n",
       "      <td>800</td>\n",
       "      <td>11</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>77</th>\n",
       "      <td>0.453022</td>\n",
       "      <td>0.010380</td>\n",
       "      <td>0.975002</td>\n",
       "      <td>41.254684</td>\n",
       "      <td>700</td>\n",
       "      <td>11</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>119</th>\n",
       "      <td>0.816264</td>\n",
       "      <td>0.007266</td>\n",
       "      <td>0.996313</td>\n",
       "      <td>24.459294</td>\n",
       "      <td>900</td>\n",
       "      <td>10</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99</th>\n",
       "      <td>0.432183</td>\n",
       "      <td>0.013997</td>\n",
       "      <td>0.987333</td>\n",
       "      <td>27.159777</td>\n",
       "      <td>900</td>\n",
       "      <td>10</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.421025</td>\n",
       "      <td>0.004999</td>\n",
       "      <td>0.999331</td>\n",
       "      <td>40.394495</td>\n",
       "      <td>800</td>\n",
       "      <td>11</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>0.814702</td>\n",
       "      <td>0.008901</td>\n",
       "      <td>0.997978</td>\n",
       "      <td>25.918714</td>\n",
       "      <td>700</td>\n",
       "      <td>11</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>0.450537</td>\n",
       "      <td>0.008609</td>\n",
       "      <td>0.999001</td>\n",
       "      <td>51.183641</td>\n",
       "      <td>800</td>\n",
       "      <td>11</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.814444</td>\n",
       "      <td>0.009408</td>\n",
       "      <td>0.988403</td>\n",
       "      <td>11.929293</td>\n",
       "      <td>700</td>\n",
       "      <td>9</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>101</th>\n",
       "      <td>0.815613</td>\n",
       "      <td>0.005455</td>\n",
       "      <td>0.999807</td>\n",
       "      <td>36.785588</td>\n",
       "      <td>800</td>\n",
       "      <td>11</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.417949</td>\n",
       "      <td>0.003396</td>\n",
       "      <td>0.999752</td>\n",
       "      <td>73.479076</td>\n",
       "      <td>800</td>\n",
       "      <td>13</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.9</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>123</th>\n",
       "      <td>0.811319</td>\n",
       "      <td>0.014001</td>\n",
       "      <td>0.996667</td>\n",
       "      <td>9.914355</td>\n",
       "      <td>800</td>\n",
       "      <td>8</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>87</th>\n",
       "      <td>0.431047</td>\n",
       "      <td>0.015503</td>\n",
       "      <td>0.942588</td>\n",
       "      <td>13.734863</td>\n",
       "      <td>700</td>\n",
       "      <td>9</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>67</th>\n",
       "      <td>0.451083</td>\n",
       "      <td>0.009110</td>\n",
       "      <td>0.900626</td>\n",
       "      <td>17.042913</td>\n",
       "      <td>700</td>\n",
       "      <td>9</td>\n",
       "      <td>0.025</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>113</th>\n",
       "      <td>0.815609</td>\n",
       "      <td>0.009512</td>\n",
       "      <td>0.998416</td>\n",
       "      <td>15.911620</td>\n",
       "      <td>800</td>\n",
       "      <td>9</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53</th>\n",
       "      <td>0.450445</td>\n",
       "      <td>0.011749</td>\n",
       "      <td>0.982936</td>\n",
       "      <td>20.165769</td>\n",
       "      <td>800</td>\n",
       "      <td>9</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     mean_test_score  std_test_score  mean_train_score  mean_fit_time  \\\n",
       "20          0.815052        0.010096          0.999442      44.528760   \n",
       "6           0.426461        0.006985          0.845973       3.892007   \n",
       "79          0.453990        0.010732          0.970382      34.947673   \n",
       "57          0.453582        0.010630          0.976138      41.552435   \n",
       "120         0.812194        0.013511          0.999468      45.108047   \n",
       "117         0.818274        0.007815          0.997005      27.323182   \n",
       "84          0.432648        0.016297          0.942675      13.140939   \n",
       "100         0.817208        0.007655          0.999240      47.867464   \n",
       "60          0.453670        0.011622          0.993961      71.353427   \n",
       "16          0.421583        0.005526          0.999723      63.789768   \n",
       "139         0.811933        0.013830          0.997609      24.026194   \n",
       "39          0.814785        0.009315          0.997608      23.571766   \n",
       "90          0.432581        0.010002          0.825493       5.935951   \n",
       "59          0.453180        0.010499          0.970333      34.490594   \n",
       "40          0.452273        0.010612          0.994166      71.622012   \n",
       "121         0.811535        0.014641          0.999844      34.210339   \n",
       "77          0.453022        0.010380          0.975002      41.254684   \n",
       "119         0.816264        0.007266          0.996313      24.459294   \n",
       "99          0.432183        0.013997          0.987333      27.159777   \n",
       "1           0.421025        0.004999          0.999331      40.394495   \n",
       "37          0.814702        0.008901          0.997978      25.918714   \n",
       "41          0.450537        0.008609          0.999001      51.183641   \n",
       "27          0.814444        0.009408          0.988403      11.929293   \n",
       "101         0.815613        0.005455          0.999807      36.785588   \n",
       "5           0.417949        0.003396          0.999752      73.479076   \n",
       "123         0.811319        0.014001          0.996667       9.914355   \n",
       "87          0.431047        0.015503          0.942588      13.734863   \n",
       "67          0.451083        0.009110          0.900626      17.042913   \n",
       "113         0.815609        0.009512          0.998416      15.911620   \n",
       "53          0.450445        0.011749          0.982936      20.165769   \n",
       "\n",
       "     param_n_estimators  param_max_depth  param_learning_rate  \\\n",
       "20                  800               12                0.025   \n",
       "6                   500                7                0.050   \n",
       "79                  900               10                0.025   \n",
       "57                  700               11                0.025   \n",
       "120                 800               12                0.025   \n",
       "117                 700               11                0.025   \n",
       "84                 1000                8                0.025   \n",
       "100                 800               12                0.025   \n",
       "60                  800               12                0.025   \n",
       "16                  700               13                0.050   \n",
       "139                 900               10                0.025   \n",
       "39                  900               10                0.025   \n",
       "90                  800                7                0.025   \n",
       "59                  900               10                0.025   \n",
       "40                  800               12                0.025   \n",
       "121                 800               11                0.050   \n",
       "77                  700               11                0.025   \n",
       "119                 900               10                0.025   \n",
       "99                  900               10                0.025   \n",
       "1                   800               11                0.050   \n",
       "37                  700               11                0.025   \n",
       "41                  800               11                0.050   \n",
       "27                  700                9                0.025   \n",
       "101                 800               11                0.050   \n",
       "5                   800               13                0.050   \n",
       "123                 800                8                0.050   \n",
       "87                  700                9                0.025   \n",
       "67                  700                9                0.025   \n",
       "113                 800                9                0.050   \n",
       "53                  800                9                0.050   \n",
       "\n",
       "     param_colsample_bytree  param_colsample_bylevel  \n",
       "20                      0.3                      0.7  \n",
       "6                       0.5                      0.5  \n",
       "79                      0.7                      0.7  \n",
       "57                      0.7                      0.7  \n",
       "120                     0.3                      0.7  \n",
       "117                     0.7                      0.7  \n",
       "84                      0.7                      0.9  \n",
       "100                     0.3                      0.7  \n",
       "60                      0.3                      0.7  \n",
       "16                      0.5                      0.9  \n",
       "139                     0.7                      0.7  \n",
       "39                      0.7                      0.7  \n",
       "90                      0.5                      0.3  \n",
       "59                      0.7                      0.7  \n",
       "40                      0.3                      0.7  \n",
       "121                     0.5                      0.3  \n",
       "77                      0.7                      0.7  \n",
       "119                     0.7                      0.7  \n",
       "99                      0.7                      0.7  \n",
       "1                       0.5                      0.3  \n",
       "37                      0.7                      0.7  \n",
       "41                      0.5                      0.3  \n",
       "27                      0.5                      0.9  \n",
       "101                     0.5                      0.3  \n",
       "5                       0.9                      0.9  \n",
       "123                     0.3                      0.7  \n",
       "87                      0.5                      0.9  \n",
       "67                      0.5                      0.9  \n",
       "113                     0.3                      0.9  \n",
       "53                      0.3                      0.9  "
      ]
     },
     "execution_count": 624,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tuning_results = os.path.join('../hyperparameter-tuning-results-20220730-00-59-29.csv')\n",
    "results = pd.read_csv(tuning_results, sep='\\t').sort_values(by='param_max_depth')\n",
    "results.sort_values(by='rank_test_score', ascending=True)[['mean_test_score', 'std_test_score', 'mean_train_score', 'mean_fit_time', 'param_n_estimators', 'param_max_depth', 'param_learning_rate',  'param_colsample_bytree', 'param_colsample_bylevel']].head(30)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "# tuning_results = os.path.join('hyperparameter-tuning-results.csv')\n",
    "tuning_results_1 = os.path.join('../hyperparameter-tuning-results-nl-sample-promising.csv')\n",
    "tuning_results_2 = os.path.join('../hyperparameter-tuning-results-nl-sample-50-small-depth.csv')\n",
    "tuning_results_3 = os.path.join('../results/hyperparameter-tuning-results-nl-sample-50-smaller-space-5fold-random-split.csv')\n",
    "tuning_results_4 = os.path.join('../results/hyperparameter-tuning-results-nl-sample-50-smaller-space-5fold-city-split.csv')\n",
    "tuning_results_5 = os.path.join('../results/hyperparameter-tuning-results-full-nl-max-depth-impact.csv')\n",
    "tuning_results_6 = os.path.join('../results/hyperparameter-tuning-results-nl-sample-10-neighborhood-classification.csv')\n",
    "\n",
    "# plot performance of distinct hyperparameter\n",
    "results = pd.read_csv(tuning_results_5, sep='\\t').sort_values(by='param_max_depth')\n",
    "visualizations.plot_hyperparameter_tuning_results(results, 'max_depth', ['score'], include_fit_time=True)\n",
    "plt.savefig('../figures/hyperparameter-tuning-max-depth.svg')\n",
    "# results.sort_values(by='rank_test_score', ascending=True)[['mean_test_score', 'std_test_score', 'mean_train_score', 'mean_fit_time', 'param_n_estimators', 'param_max_depth', 'param_learning_rate',  'param_colsample_bytree', 'param_colsample_bylevel']]#.head(30)\n",
    "\n",
    "# plot mean performance of parameters during random search\n",
    "# results = pd.concat([pd.read_csv(path, sep='\\t').sort_values(by='param_n_estimators') for path in [tuning_results_3, tuning_results_4]], ignore_index=True)\n",
    "# results = results.sort_values(by='rank_test_score', ascending=True).head(50).groupby('param_n_estimators', as_index=False).mean()\n",
    "# visualizations.plot_hyperparameter_tuning_results(results, 'n_estimators', ['score'])\n",
    "\n",
    "\n",
    "# plt.savefig('../figures/hyperparameter-tuning.svg')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prediction Examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 697,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-15 17:32:56,718 | INFO : Dataset length: 176319\n",
      "2022-07-15 17:32:56,764 | INFO : Dataset allocated memory: 184 MB\n",
      "2022-07-15 17:32:57,120 | INFO : Dataset standard deviation: 28.072581085899984\n",
      "2022-07-15 17:32:57,122 | INFO : Dataset mean age: 1974.2587129010487\n",
      "2022-07-15 17:32:57,123 | INFO : Training dataset length: 141055\n",
      "2022-07-15 17:32:57,131 | INFO : Test dataset length: 35264\n",
      "2022-07-15 17:32:57,611 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-07-15 17:32:59,292 | WARNING : Building type information missing. Considering all buildings as residential.\n",
      "2022-07-15 17:33:00,838 | INFO : Test dataset standard deviation after preprocessing: 25.07194896229578\n",
      "2022-07-15 17:33:00,840 | INFO : Test dataset mean age after preprocessing: 1975.4969008264463\n",
      "2022-07-15 17:33:00,841 | INFO : Training dataset length after preprocessing: 139362\n",
      "2022-07-15 17:33:00,842 | INFO : Test dataset length after preprocessing: 34848\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-15 17:33:05,344 | ERROR : Failed to calculate energy error: [Errno 2] No such file or directory: '/p/projects/eubucco/data/3-ml-inputs/TABULA_heating_demand.csv'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE: 10.92 y\n",
      "RMSE: 17.01 y\n",
      "R2: 0.5396\n",
      "R2: nan\n",
      "MAPE: nan\n"
     ]
    }
   ],
   "source": [
    "\n",
    "predictor = AgePredictor(\n",
    "    model=XGBRegressor(tree_method='hist', n_jobs=-1),\n",
    "    df=os.path.join(PROJECT_SRC_PATH, '..', 'data', 'exp', 'df-NLD-exp.pkl'),\n",
    "    test_training_split=split_80_20,\n",
    "    # cross_validation_split=city_cross_validation,\n",
    "    preprocessing_stages=[remove_outliers, preparation.add_residential_type_column],\n",
    ")\n",
    "\n",
    "predictor.print_model_error()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 701,
   "metadata": {},
   "outputs": [
    {
     "ename": "AttributeError",
     "evalue": "'XGBRegressor' object has no attribute '__del__'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mAttributeError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[1;32m/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb Cell 36'\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000146?line=0'>1</a>\u001b[0m predictor\u001b[39m.\u001b[39;49mmodel\u001b[39m.\u001b[39;49m\u001b[39m__del__\u001b[39;49m()\n",
      "\u001b[0;31mAttributeError\u001b[0m: 'XGBRegressor' object has no attribute '__del__'"
     ]
    }
   ],
   "source": [
    "predictor.model.__del__()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-07 13:55:31,898 | INFO : Generated bins: [0, 1915, 1949, 1968, 1975, 1982, 1990, 2000, 2006, 2013, 2051]\n",
      "2022-07-07 13:55:31,899 | INFO : Generated bins with the following labels: ['<1915', '1915-1948', '1949-1967', '1968-1974', '1975-1981', '1982-1989', '1990-1999', '2000-2005', '2006-2012', '2013-2050']\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification report:\n",
      "               precision    recall  f1-score  support\n",
      "<1915          0.343567  0.321038  0.331921      732\n",
      "1915-1948      0.493369  0.532037  0.511974     1748\n",
      "1949-1967      0.533437  0.558087  0.545483     3073\n",
      "1968-1974      0.392661  0.412129  0.402160     2259\n",
      "1975-1981      0.473380  0.507999  0.490079     3063\n",
      "1982-1989      0.426824  0.430687  0.428747     2431\n",
      "1990-1999      0.515502  0.476026  0.494978     2899\n",
      "2000-2005      0.474227  0.435194  0.453873     2114\n",
      "2006-2012      0.534068  0.523600  0.528782     2500\n",
      "2013-2050      0.357504  0.287653  0.318797      737\n",
      "accuracy       0.474810  0.474810  0.474810        0\n",
      "macro avg      0.454454  0.448445  0.450679    21556\n",
      "weighted avg   0.474269  0.474810  0.473992    21556\n",
      "Cohen’s kappa: 0.4065\n",
      "Matthews correlation coefficient (MCC): 0.4066\n",
      "R2: 0.1144\n",
      "MAPE: 0.4566\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    }
   ],
   "source": [
    "\n",
    "equally_sized_bins = (1900, 2020, 10)\n",
    "\n",
    "predictor = AgeClassifier(\n",
    "    model=XGBClassifier(tree_method='hist', n_jobs=-1),\n",
    "    df_path=os.path.join(PROJECT_SRC_PATH, '..', 'data', 'exp', 'df-FRA-exp.pkl'),\n",
    "    # test_training_split=split_80_20,\n",
    "    cross_validation_split=cross_validation,\n",
    "    preprocessing_stages=[remove_outliers, preparation.add_residential_type_column],\n",
    "    mitigate_class_imbalance=True,\n",
    "    bins=dataset.TABULA_AGE_BINS['france'],\n",
    ")\n",
    "\n",
    "predictor.print_model_error()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "      <th>support</th>\n",
       "      <th>kappa</th>\n",
       "      <th>mcc</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1900-1909</th>\n",
       "      <td>0.155945</td>\n",
       "      <td>0.314961</td>\n",
       "      <td>0.208605</td>\n",
       "      <td>254.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1910-1919</th>\n",
       "      <td>0.167610</td>\n",
       "      <td>0.307054</td>\n",
       "      <td>0.216850</td>\n",
       "      <td>482.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1920-1929</th>\n",
       "      <td>0.211898</td>\n",
       "      <td>0.419283</td>\n",
       "      <td>0.281521</td>\n",
       "      <td>892.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1930-1939</th>\n",
       "      <td>0.315101</td>\n",
       "      <td>0.473315</td>\n",
       "      <td>0.378333</td>\n",
       "      <td>1424.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1940-1949</th>\n",
       "      <td>0.277852</td>\n",
       "      <td>0.360000</td>\n",
       "      <td>0.313636</td>\n",
       "      <td>575.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1950-1959</th>\n",
       "      <td>0.484458</td>\n",
       "      <td>0.536909</td>\n",
       "      <td>0.509337</td>\n",
       "      <td>2032.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1960-1969</th>\n",
       "      <td>0.650462</td>\n",
       "      <td>0.587059</td>\n",
       "      <td>0.617136</td>\n",
       "      <td>3601.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1970-1979</th>\n",
       "      <td>0.774424</td>\n",
       "      <td>0.629637</td>\n",
       "      <td>0.694565</td>\n",
       "      <td>5338.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1980-1989</th>\n",
       "      <td>0.769835</td>\n",
       "      <td>0.612400</td>\n",
       "      <td>0.682151</td>\n",
       "      <td>4484.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1990-1999</th>\n",
       "      <td>0.699196</td>\n",
       "      <td>0.595232</td>\n",
       "      <td>0.643039</td>\n",
       "      <td>4237.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2000-2009</th>\n",
       "      <td>0.617001</td>\n",
       "      <td>0.545301</td>\n",
       "      <td>0.578940</td>\n",
       "      <td>3554.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-2019</th>\n",
       "      <td>0.399919</td>\n",
       "      <td>0.548835</td>\n",
       "      <td>0.462690</td>\n",
       "      <td>1802.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>macro avg</th>\n",
       "      <td>0.460308</td>\n",
       "      <td>0.494165</td>\n",
       "      <td>0.465567</td>\n",
       "      <td>28675.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>weighted avg</th>\n",
       "      <td>0.617485</td>\n",
       "      <td>0.566486</td>\n",
       "      <td>0.584733</td>\n",
       "      <td>28675.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>total</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.511906</td>\n",
       "      <td>0.513545</td>\n",
       "      <td>0.566486</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              precision    recall  f1-score  support     kappa       mcc  \\\n",
       "1900-1909      0.155945  0.314961  0.208605    254.0       NaN       NaN   \n",
       "1910-1919      0.167610  0.307054  0.216850    482.0       NaN       NaN   \n",
       "1920-1929      0.211898  0.419283  0.281521    892.0       NaN       NaN   \n",
       "1930-1939      0.315101  0.473315  0.378333   1424.0       NaN       NaN   \n",
       "1940-1949      0.277852  0.360000  0.313636    575.0       NaN       NaN   \n",
       "1950-1959      0.484458  0.536909  0.509337   2032.0       NaN       NaN   \n",
       "1960-1969      0.650462  0.587059  0.617136   3601.0       NaN       NaN   \n",
       "1970-1979      0.774424  0.629637  0.694565   5338.0       NaN       NaN   \n",
       "1980-1989      0.769835  0.612400  0.682151   4484.0       NaN       NaN   \n",
       "1990-1999      0.699196  0.595232  0.643039   4237.0       NaN       NaN   \n",
       "2000-2009      0.617001  0.545301  0.578940   3554.0       NaN       NaN   \n",
       "2010-2019      0.399919  0.548835  0.462690   1802.0       NaN       NaN   \n",
       "macro avg      0.460308  0.494165  0.465567  28675.0       NaN       NaN   \n",
       "weighted avg   0.617485  0.566486  0.584733  28675.0       NaN       NaN   \n",
       "total               NaN       NaN       NaN      NaN  0.511906  0.513545   \n",
       "\n",
       "              accuracy  \n",
       "1900-1909          NaN  \n",
       "1910-1919          NaN  \n",
       "1920-1929          NaN  \n",
       "1930-1939          NaN  \n",
       "1940-1949          NaN  \n",
       "1950-1959          NaN  \n",
       "1960-1969          NaN  \n",
       "1970-1979          NaN  \n",
       "1980-1989          NaN  \n",
       "1990-1999          NaN  \n",
       "2000-2009          NaN  \n",
       "2010-2019          NaN  \n",
       "macro avg          NaN  \n",
       "weighted avg       NaN  \n",
       "total         0.566486  "
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# predictor.classification_report()\n",
    "predictor.save_model_error()\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Main experiments - local test run"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 489,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[autoreload of prediction_age failed: Traceback (most recent call last):\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 257, in check\n",
      "    superreload(m, reload, self.old_objects)\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 480, in superreload\n",
      "    update_generic(old_obj, new_obj)\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 377, in update_generic\n",
      "    update(a, b)\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 329, in update_class\n",
      "    if update_generic(old_obj, new_obj):\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 377, in update_generic\n",
      "    update(a, b)\n",
      "  File \"/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/IPython/extensions/autoreload.py\", line 289, in update_function\n",
      "    setattr(old, name, getattr(new, name))\n",
      "ValueError: mcc() requires a code object with 0 free vars, not 1\n",
      "]\n",
      "2022-07-29 11:10:55,577 | INFO : RQ2 - Evaluating prediction performance differences for FR, ES & NL...\n",
      "2022-07-29 11:10:55,582 | INFO : Comparing regression models...\n",
      "2022-07-29 11:10:55,866 | INFO : Dataset standard deviation: 64.31994496480065\n",
      "2022-07-29 11:10:55,868 | INFO : Dataset mean age: 1950.0572808833679\n",
      "2022-07-29 11:10:55,868 | INFO : Training dataset length: 39412\n",
      "2022-07-29 11:10:55,869 | INFO : Test dataset length: 9854\n",
      "2022-07-29 11:10:56,016 | INFO : Test dataset standard deviation after preprocessing: 31.02043259805174\n",
      "2022-07-29 11:10:56,018 | INFO : Test dataset mean age after preprocessing: 1973.857247706422\n",
      "2022-07-29 11:10:56,018 | INFO : Training dataset length after preprocessing: 32780\n",
      "2022-07-29 11:10:56,019 | INFO : Test dataset length after preprocessing: 8175\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:10:56,991 | INFO : Dataset standard deviation: 64.31994496480065\n",
      "2022-07-29 11:10:56,993 | INFO : Dataset mean age: 1950.0572808833679\n",
      "2022-07-29 11:10:56,993 | INFO : Training dataset length: 39413\n",
      "2022-07-29 11:10:56,994 | INFO : Test dataset length: 9853\n",
      "2022-07-29 11:10:57,131 | INFO : Test dataset standard deviation after preprocessing: 31.399171315446154\n",
      "2022-07-29 11:10:57,132 | INFO : Test dataset mean age after preprocessing: 1974.0201424711372\n",
      "2022-07-29 11:10:57,133 | INFO : Training dataset length after preprocessing: 32813\n",
      "2022-07-29 11:10:57,133 | INFO : Test dataset length after preprocessing: 8142\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:10:58,047 | INFO : Dataset standard deviation: 64.31994496480065\n",
      "2022-07-29 11:10:58,048 | INFO : Dataset mean age: 1950.0572808833679\n",
      "2022-07-29 11:10:58,049 | INFO : Training dataset length: 39413\n",
      "2022-07-29 11:10:58,049 | INFO : Test dataset length: 9853\n",
      "2022-07-29 11:10:58,180 | INFO : Test dataset standard deviation after preprocessing: 31.062605484453126\n",
      "2022-07-29 11:10:58,181 | INFO : Test dataset mean age after preprocessing: 1973.837055214724\n",
      "2022-07-29 11:10:58,182 | INFO : Training dataset length after preprocessing: 32805\n",
      "2022-07-29 11:10:58,183 | INFO : Test dataset length after preprocessing: 8150\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:10:59,075 | INFO : Dataset standard deviation: 64.31994496480065\n",
      "2022-07-29 11:10:59,078 | INFO : Dataset mean age: 1950.0572808833679\n",
      "2022-07-29 11:10:59,079 | INFO : Training dataset length: 39413\n",
      "2022-07-29 11:10:59,079 | INFO : Test dataset length: 9853\n",
      "2022-07-29 11:10:59,218 | INFO : Test dataset standard deviation after preprocessing: 31.34413512074983\n",
      "2022-07-29 11:10:59,219 | INFO : Test dataset mean age after preprocessing: 1973.319859615152\n",
      "2022-07-29 11:10:59,220 | INFO : Training dataset length after preprocessing: 32692\n",
      "2022-07-29 11:10:59,220 | INFO : Test dataset length after preprocessing: 8263\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:11:00,111 | INFO : Dataset standard deviation: 64.31994496480065\n",
      "2022-07-29 11:11:00,113 | INFO : Dataset mean age: 1950.0572808833679\n",
      "2022-07-29 11:11:00,113 | INFO : Training dataset length: 39413\n",
      "2022-07-29 11:11:00,114 | INFO : Test dataset length: 9853\n",
      "2022-07-29 11:11:00,249 | INFO : Test dataset standard deviation after preprocessing: 30.56128709160291\n",
      "2022-07-29 11:11:00,250 | INFO : Test dataset mean age after preprocessing: 1974.1692401215805\n",
      "2022-07-29 11:11:00,250 | INFO : Training dataset length after preprocessing: 32730\n",
      "2022-07-29 11:11:00,251 | INFO : Test dataset length after preprocessing: 8225\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:11:04,277 | INFO : Dataset standard deviation: 34.886836696402796\n",
      "2022-07-29 11:11:04,278 | INFO : Dataset mean age: 1980.367768406656\n",
      "2022-07-29 11:11:04,279 | INFO : Training dataset length: 17548\n",
      "2022-07-29 11:11:04,280 | INFO : Test dataset length: 4387\n",
      "2022-07-29 11:11:04,362 | INFO : Test dataset standard deviation after preprocessing: 26.760150029573992\n",
      "2022-07-29 11:11:04,363 | INFO : Test dataset mean age after preprocessing: 1980.7655963302752\n",
      "2022-07-29 11:11:04,363 | INFO : Training dataset length after preprocessing: 17447\n",
      "2022-07-29 11:11:04,364 | INFO : Test dataset length after preprocessing: 4360\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:11:05,019 | INFO : Dataset standard deviation: 34.886836696402796\n",
      "2022-07-29 11:11:05,022 | INFO : Dataset mean age: 1980.367768406656\n",
      "2022-07-29 11:11:05,022 | INFO : Training dataset length: 17548\n",
      "2022-07-29 11:11:05,023 | INFO : Test dataset length: 4387\n",
      "2022-07-29 11:11:05,092 | INFO : Test dataset standard deviation after preprocessing: 26.405115500513716\n",
      "2022-07-29 11:11:05,094 | INFO : Test dataset mean age after preprocessing: 1981.7741861531408\n",
      "2022-07-29 11:11:05,094 | INFO : Training dataset length after preprocessing: 17445\n",
      "2022-07-29 11:11:05,095 | INFO : Test dataset length after preprocessing: 4362\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:11:05,748 | INFO : Dataset standard deviation: 34.886836696402796\n",
      "2022-07-29 11:11:05,750 | INFO : Dataset mean age: 1980.367768406656\n",
      "2022-07-29 11:11:05,750 | INFO : Training dataset length: 17548\n",
      "2022-07-29 11:11:05,751 | INFO : Test dataset length: 4387\n",
      "2022-07-29 11:11:05,820 | INFO : Test dataset standard deviation after preprocessing: 26.960451748809625\n",
      "2022-07-29 11:11:05,821 | INFO : Test dataset mean age after preprocessing: 1981.6158788435062\n",
      "2022-07-29 11:11:05,822 | INFO : Training dataset length after preprocessing: 17449\n",
      "2022-07-29 11:11:05,823 | INFO : Test dataset length after preprocessing: 4358\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:11:06,483 | INFO : Dataset standard deviation: 34.886836696402796\n",
      "2022-07-29 11:11:06,484 | INFO : Dataset mean age: 1980.367768406656\n",
      "2022-07-29 11:11:06,485 | INFO : Training dataset length: 17548\n",
      "2022-07-29 11:11:06,486 | INFO : Test dataset length: 4387\n",
      "2022-07-29 11:11:06,555 | INFO : Test dataset standard deviation after preprocessing: 26.657725482585082\n",
      "2022-07-29 11:11:06,556 | INFO : Test dataset mean age after preprocessing: 1980.708419362239\n",
      "2022-07-29 11:11:06,556 | INFO : Training dataset length after preprocessing: 17448\n",
      "2022-07-29 11:11:06,557 | INFO : Test dataset length after preprocessing: 4359\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:11:07,280 | INFO : Dataset standard deviation: 34.886836696402796\n",
      "2022-07-29 11:11:07,281 | INFO : Dataset mean age: 1980.367768406656\n",
      "2022-07-29 11:11:07,282 | INFO : Training dataset length: 17548\n",
      "2022-07-29 11:11:07,282 | INFO : Test dataset length: 4387\n",
      "2022-07-29 11:11:07,354 | INFO : Test dataset standard deviation after preprocessing: 26.729158810000296\n",
      "2022-07-29 11:11:07,355 | INFO : Test dataset mean age after preprocessing: 1981.7987637362637\n",
      "2022-07-29 11:11:07,356 | INFO : Training dataset length after preprocessing: 17439\n",
      "2022-07-29 11:11:07,357 | INFO : Test dataset length after preprocessing: 4368\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-29 11:11:09,442 | INFO : Saving comparison metrics...\n",
      "2022-07-29 11:11:09,450 | INFO :                name        R2  R2_std        MAE  MAE_std       RMSE  \\\n",
      "0  France_random-cv  0.410915     0.0  17.589786      0.0  23.853014   \n",
      "1   Spain_random-cv  0.507240     0.0  14.110243      0.0  18.745747   \n",
      "\n",
      "   RMSE_std  within_5_years  within_10_years  within_20_years  R2_seed_0  \\\n",
      "0       0.0        0.210866         0.406275         0.683580   0.410915   \n",
      "1       0.0        0.196176         0.437337         0.810978   0.507240   \n",
      "\n",
      "      MCC_5  MCC_std_5    MCC_10  MCC_std_10    MCC_20  MCC_std_20  \n",
      "0  0.052963        0.0  0.112499         0.0  0.193589         0.0  \n",
      "1  0.034009        0.0  0.084466         0.0  0.190392         0.0  \n"
     ]
    }
   ],
   "source": [
    "\n",
    "PROJECT_BIN_PATH = os.path.join(os.path.abspath(''), '..', 'bin')\n",
    "sys.path.append(PROJECT_BIN_PATH)\n",
    "import run_experiments as exp\n",
    "\n",
    "logging.getLogger('prediction').setLevel(logging.WARNING)\n",
    "\n",
    "method = 'regression'\n",
    "\n",
    "exp.FRAC = None\n",
    "# exp.DATA_DIR = os.path.join(PROJECT_BIN_PATH, '..', 'data', 'exp')\n",
    "# exp.RESULT_DIR = os.path.join(PROJECT_BIN_PATH, '..', 'data', 'exp', 'results')\n",
    "exp.XGBOOST_PARAMS = {'tree_method': 'hist'}\n",
    "exp.HYPERPARAMETERS = {\n",
    "    'max_depth': 5,\n",
    "    'learning_rate': 0.1,\n",
    "    'n_estimators': 100,\n",
    "    'colsample_bytree': 0.9,\n",
    "    'colsample_bylevel': 0.5,\n",
    "}\n",
    "\n",
    "# exp.compare_energy_error()\n",
    "exp.compare_countries(method)\n",
    "# exp.generalize_across_countries(method)\n",
    "# exp.evaluate_impact_of_additional_data(method, across_cities=True)\n",
    "# exp.evaluate_impact_of_spatial_distance_on_generalization(method)\n",
    "# exp.evaluate_impact_of_additional_data(method, across_cities=True, include_data_from_other_countries=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.4849    15\n",
       "2.5366    15\n",
       "2.6051    14\n",
       "2.5943    13\n",
       "2.4877    13\n",
       "          ..\n",
       "4.7236     1\n",
       "7.0996     1\n",
       "7.8252     1\n",
       "7.8522     1\n",
       "3.6177     1\n",
       "Name: height, Length: 130494, dtype: int64"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_nl_sample['height'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Main experiment - large-scale results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <th>Model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Spain</th>\n",
       "      <th>neighborhood-cv</th>\n",
       "      <td>Spain_neighborhood-cv</td>\n",
       "      <td>19.3</td>\n",
       "      <td>26.3</td>\n",
       "      <td>0.34</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <th>neighborhood-cv</th>\n",
       "      <td>All_neighborhood-cv</td>\n",
       "      <td>17.8</td>\n",
       "      <td>24.3</td>\n",
       "      <td>0.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>France</th>\n",
       "      <th>neighborhood-cv</th>\n",
       "      <td>France_neighborhood-cv</td>\n",
       "      <td>18.0</td>\n",
       "      <td>24.4</td>\n",
       "      <td>0.41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Spain</th>\n",
       "      <th>random-cv</th>\n",
       "      <td>Spain_random-cv</td>\n",
       "      <td>17.5</td>\n",
       "      <td>24.5</td>\n",
       "      <td>0.43</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Netherlands</th>\n",
       "      <th>neighborhood-cv</th>\n",
       "      <td>Netherlands_neighborhood-cv</td>\n",
       "      <td>14.3</td>\n",
       "      <td>19.9</td>\n",
       "      <td>0.44</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <th>random-cv</th>\n",
       "      <td>All_random-cv</td>\n",
       "      <td>16.2</td>\n",
       "      <td>22.9</td>\n",
       "      <td>0.45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>France</th>\n",
       "      <th>random-cv</th>\n",
       "      <td>France_random-cv</td>\n",
       "      <td>16.8</td>\n",
       "      <td>23.4</td>\n",
       "      <td>0.46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Netherlands</th>\n",
       "      <th>random-cv</th>\n",
       "      <td>Netherlands_random-cv</td>\n",
       "      <td>10.1</td>\n",
       "      <td>16.1</td>\n",
       "      <td>0.63</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Spain</th>\n",
       "      <th>city-cv</th>\n",
       "      <td>Spain_city-cv</td>\n",
       "      <td>19.9</td>\n",
       "      <td>26.9</td>\n",
       "      <td>0.32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <th>city-cv</th>\n",
       "      <td>All_city-cv</td>\n",
       "      <td>18.3</td>\n",
       "      <td>24.8</td>\n",
       "      <td>0.36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Netherlands</th>\n",
       "      <th>city-cv</th>\n",
       "      <td>Netherlands_city-cv</td>\n",
       "      <td>15.3</td>\n",
       "      <td>20.9</td>\n",
       "      <td>0.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>France</th>\n",
       "      <th>city-cv</th>\n",
       "      <td>France_city-cv</td>\n",
       "      <td>18.3</td>\n",
       "      <td>24.7</td>\n",
       "      <td>0.39</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <th>country-cv</th>\n",
       "      <td>All_country-cv</td>\n",
       "      <td>25.3</td>\n",
       "      <td>31.2</td>\n",
       "      <td>-0.02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>France</th>\n",
       "      <th>country-cv</th>\n",
       "      <td>France_country-cv</td>\n",
       "      <td>23.8</td>\n",
       "      <td>29.5</td>\n",
       "      <td>0.13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Spain</th>\n",
       "      <th>country-cv</th>\n",
       "      <td>Spain_country-cv</td>\n",
       "      <td>28.6</td>\n",
       "      <td>34.7</td>\n",
       "      <td>-0.15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Netherlands</th>\n",
       "      <th>country-cv</th>\n",
       "      <td>Netherlands_country-cv</td>\n",
       "      <td>23.4</td>\n",
       "      <td>28.8</td>\n",
       "      <td>-0.18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Spain</th>\n",
       "      <th>block-cv</th>\n",
       "      <td>Spain_block-cv</td>\n",
       "      <td>18.5</td>\n",
       "      <td>25.5</td>\n",
       "      <td>0.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <th>block-cv</th>\n",
       "      <td>All_block-cv</td>\n",
       "      <td>17.0</td>\n",
       "      <td>23.6</td>\n",
       "      <td>0.41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>France</th>\n",
       "      <th>block-cv</th>\n",
       "      <td>France_block-cv</td>\n",
       "      <td>17.6</td>\n",
       "      <td>24.1</td>\n",
       "      <td>0.42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Netherlands</th>\n",
       "      <th>block-cv</th>\n",
       "      <td>Netherlands_block-cv</td>\n",
       "      <td>12.5</td>\n",
       "      <td>18.5</td>\n",
       "      <td>0.51</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    name   MAE  RMSE    R2\n",
       "Country     Model                                                         \n",
       "Spain       neighborhood-cv        Spain_neighborhood-cv  19.3  26.3  0.34\n",
       "All         neighborhood-cv          All_neighborhood-cv  17.8  24.3  0.38\n",
       "France      neighborhood-cv       France_neighborhood-cv  18.0  24.4  0.41\n",
       "Spain       random-cv                    Spain_random-cv  17.5  24.5  0.43\n",
       "Netherlands neighborhood-cv  Netherlands_neighborhood-cv  14.3  19.9  0.44\n",
       "All         random-cv                      All_random-cv  16.2  22.9  0.45\n",
       "France      random-cv                   France_random-cv  16.8  23.4  0.46\n",
       "Netherlands random-cv              Netherlands_random-cv  10.1  16.1  0.63\n",
       "Spain       city-cv                        Spain_city-cv  19.9  26.9  0.32\n",
       "All         city-cv                          All_city-cv  18.3  24.8  0.36\n",
       "Netherlands city-cv                  Netherlands_city-cv  15.3  20.9  0.38\n",
       "France      city-cv                       France_city-cv  18.3  24.7  0.39\n",
       "All         country-cv                    All_country-cv  25.3  31.2 -0.02\n",
       "France      country-cv                 France_country-cv  23.8  29.5  0.13\n",
       "Spain       country-cv                  Spain_country-cv  28.6  34.7 -0.15\n",
       "Netherlands country-cv            Netherlands_country-cv  23.4  28.8 -0.18\n",
       "Spain       block-cv                      Spain_block-cv  18.5  25.5  0.38\n",
       "All         block-cv                        All_block-cv  17.0  23.6  0.41\n",
       "France      block-cv                     France_block-cv  17.6  24.1  0.42\n",
       "Netherlands block-cv                Netherlands_block-cv  12.5  18.5  0.51"
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg_results = pd.read_csv('../results/error-comparison-regression.csv')[['name', 'MAE', 'RMSE', 'R2']]\n",
    "# r[['MAE', 'RMSE']] = r[['MAE', 'RMSE']].round(decimals=2)\n",
    "# r['R2'] = r['R2'].round(decimals=4)\n",
    "# r\n",
    "reg_results\n",
    "\n",
    "reg_results['Country'] = reg_results['name'].str.split('_').str[0]\n",
    "reg_results['Model'] = reg_results['name'].str.split('_').str[1]\n",
    "reg_results = reg_results.set_index(['Country', 'Model'])\n",
    "reg_results['R2'] = reg_results['R2'].round(2)\n",
    "reg_results[['MAE', 'RMSE']] = reg_results[['MAE', 'RMSE']].round(1)\n",
    "reg_results.sort_index(level=1)[['MAE', 'RMSE', 'R2']].to_latex('reg-results.csv')\n",
    "reg_results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>MCC</th>\n",
       "      <th>MCC_std</th>\n",
       "      <th>F1</th>\n",
       "      <th>F1_std</th>\n",
       "      <th>Recall_1900-1909</th>\n",
       "      <th>Recall_1910-1919</th>\n",
       "      <th>Recall_1920-1929</th>\n",
       "      <th>Recall_1930-1939</th>\n",
       "      <th>Recall_1940-1949</th>\n",
       "      <th>Recall_1950-1959</th>\n",
       "      <th>Recall_1960-1969</th>\n",
       "      <th>Recall_1970-1979</th>\n",
       "      <th>Recall_1980-1989</th>\n",
       "      <th>Recall_1990-1999</th>\n",
       "      <th>Recall_2000-2009</th>\n",
       "      <th>Recall_2010-2019</th>\n",
       "      <th>MCC_seed_0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Spain_city-cv</td>\n",
       "      <td>0.14</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.19</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.44</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.08</td>\n",
       "      <td>0.10</td>\n",
       "      <td>0.16</td>\n",
       "      <td>0.12</td>\n",
       "      <td>0.24</td>\n",
       "      <td>0.19</td>\n",
       "      <td>0.16</td>\n",
       "      <td>0.17</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.27</td>\n",
       "      <td>0.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>Spain_neighborhood-cv</td>\n",
       "      <td>0.18</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.23</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.06</td>\n",
       "      <td>0.14</td>\n",
       "      <td>0.14</td>\n",
       "      <td>0.20</td>\n",
       "      <td>0.18</td>\n",
       "      <td>0.27</td>\n",
       "      <td>0.22</td>\n",
       "      <td>0.20</td>\n",
       "      <td>0.20</td>\n",
       "      <td>0.41</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>France_city-cv</td>\n",
       "      <td>0.24</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.27</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.61</td>\n",
       "      <td>0.03</td>\n",
       "      <td>0.30</td>\n",
       "      <td>0.23</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.27</td>\n",
       "      <td>0.28</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.30</td>\n",
       "      <td>0.21</td>\n",
       "      <td>0.34</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Spain_block-cv</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.64</td>\n",
       "      <td>0.24</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.32</td>\n",
       "      <td>0.34</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.28</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.24</td>\n",
       "      <td>0.40</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>France_neighborhood-cv</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.28</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.62</td>\n",
       "      <td>0.05</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.02</td>\n",
       "      <td>0.30</td>\n",
       "      <td>0.29</td>\n",
       "      <td>0.36</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.23</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.43</td>\n",
       "      <td>0.26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Netherlands_city-cv</td>\n",
       "      <td>0.28</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.29</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.20</td>\n",
       "      <td>0.06</td>\n",
       "      <td>0.16</td>\n",
       "      <td>0.37</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.46</td>\n",
       "      <td>0.47</td>\n",
       "      <td>0.41</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.36</td>\n",
       "      <td>0.37</td>\n",
       "      <td>0.28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>France_block-cv</td>\n",
       "      <td>0.30</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.61</td>\n",
       "      <td>0.12</td>\n",
       "      <td>0.39</td>\n",
       "      <td>0.29</td>\n",
       "      <td>0.09</td>\n",
       "      <td>0.37</td>\n",
       "      <td>0.34</td>\n",
       "      <td>0.40</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.27</td>\n",
       "      <td>0.39</td>\n",
       "      <td>0.45</td>\n",
       "      <td>0.30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>Netherlands_neighborhood-cv</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.32</td>\n",
       "      <td>0.10</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.03</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.45</td>\n",
       "      <td>0.38</td>\n",
       "      <td>0.38</td>\n",
       "      <td>0.41</td>\n",
       "      <td>0.33</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Spain_random-cv</td>\n",
       "      <td>0.39</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.44</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.71</td>\n",
       "      <td>0.47</td>\n",
       "      <td>0.52</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.47</td>\n",
       "      <td>0.48</td>\n",
       "      <td>0.41</td>\n",
       "      <td>0.38</td>\n",
       "      <td>0.36</td>\n",
       "      <td>0.46</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.39</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>France_random-cv</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.45</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.64</td>\n",
       "      <td>0.30</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.46</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.47</td>\n",
       "      <td>0.41</td>\n",
       "      <td>0.48</td>\n",
       "      <td>0.54</td>\n",
       "      <td>0.42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Netherlands_block-cv</td>\n",
       "      <td>0.52</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.51</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.43</td>\n",
       "      <td>0.32</td>\n",
       "      <td>0.43</td>\n",
       "      <td>0.54</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.56</td>\n",
       "      <td>0.63</td>\n",
       "      <td>0.64</td>\n",
       "      <td>0.63</td>\n",
       "      <td>0.57</td>\n",
       "      <td>0.53</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>Netherlands_random-cv</td>\n",
       "      <td>0.75</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.73</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.66</td>\n",
       "      <td>0.70</td>\n",
       "      <td>0.74</td>\n",
       "      <td>0.74</td>\n",
       "      <td>0.78</td>\n",
       "      <td>0.81</td>\n",
       "      <td>0.82</td>\n",
       "      <td>0.81</td>\n",
       "      <td>0.78</td>\n",
       "      <td>0.73</td>\n",
       "      <td>0.69</td>\n",
       "      <td>0.75</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                           name   MCC  MCC_std    F1  F1_std  \\\n",
       "0                 Spain_city-cv  0.14      0.0  0.19     0.0   \n",
       "9         Spain_neighborhood-cv  0.18      0.0  0.23     0.0   \n",
       "1                France_city-cv  0.24      0.0  0.27     0.0   \n",
       "2                Spain_block-cv  0.26      0.0  0.31     0.0   \n",
       "10       France_neighborhood-cv  0.26      0.0  0.28     0.0   \n",
       "3           Netherlands_city-cv  0.28      0.0  0.29     0.0   \n",
       "4               France_block-cv  0.30      0.0  0.33     0.0   \n",
       "11  Netherlands_neighborhood-cv  0.33      0.0  0.33     0.0   \n",
       "5               Spain_random-cv  0.39      0.0  0.44     0.0   \n",
       "6              France_random-cv  0.42      0.0  0.45     0.0   \n",
       "7          Netherlands_block-cv  0.52      0.0  0.51     0.0   \n",
       "8         Netherlands_random-cv  0.75      0.0  0.73     0.0   \n",
       "\n",
       "    Recall_1900-1909  Recall_1910-1919  Recall_1920-1929  Recall_1930-1939  \\\n",
       "0               0.44              0.01              0.08              0.10   \n",
       "9               0.50              0.06              0.14              0.14   \n",
       "1               0.61              0.03              0.30              0.23   \n",
       "2               0.64              0.24              0.35              0.32   \n",
       "10              0.62              0.05              0.35              0.26   \n",
       "3               0.20              0.06              0.16              0.37   \n",
       "4               0.61              0.12              0.39              0.29   \n",
       "11              0.32              0.10              0.26              0.42   \n",
       "5               0.71              0.47              0.52              0.49   \n",
       "6               0.64              0.30              0.49              0.42   \n",
       "7               0.43              0.32              0.43              0.54   \n",
       "8               0.68              0.66              0.70              0.74   \n",
       "\n",
       "    Recall_1940-1949  Recall_1950-1959  Recall_1960-1969  Recall_1970-1979  \\\n",
       "0               0.16              0.12              0.24              0.19   \n",
       "9               0.20              0.18              0.27              0.22   \n",
       "1               0.01              0.27              0.28              0.35   \n",
       "2               0.34              0.31              0.35              0.28   \n",
       "10              0.02              0.30              0.29              0.36   \n",
       "3               0.01              0.31              0.46              0.47   \n",
       "4               0.09              0.37              0.34              0.40   \n",
       "11              0.03              0.35              0.49              0.49   \n",
       "5               0.49              0.47              0.48              0.41   \n",
       "6               0.26              0.50              0.46              0.50   \n",
       "7               0.35              0.56              0.63              0.64   \n",
       "8               0.74              0.78              0.81              0.82   \n",
       "\n",
       "    Recall_1980-1989  Recall_1990-1999  Recall_2000-2009  Recall_2010-2019  \\\n",
       "0               0.16              0.17              0.42              0.27   \n",
       "9               0.20              0.20              0.41              0.31   \n",
       "1               0.30              0.21              0.34              0.42   \n",
       "2               0.25              0.24              0.40              0.35   \n",
       "10              0.31              0.23              0.35              0.43   \n",
       "3               0.41              0.33              0.36              0.37   \n",
       "4               0.35              0.27              0.39              0.45   \n",
       "11              0.45              0.38              0.38              0.41   \n",
       "5               0.38              0.36              0.46              0.50   \n",
       "6               0.47              0.41              0.48              0.54   \n",
       "7               0.63              0.57              0.53              0.50   \n",
       "8               0.81              0.78              0.73              0.69   \n",
       "\n",
       "    MCC_seed_0  \n",
       "0         0.14  \n",
       "9         0.18  \n",
       "1         0.24  \n",
       "2         0.26  \n",
       "10        0.26  \n",
       "3         0.28  \n",
       "4         0.30  \n",
       "11        0.33  \n",
       "5         0.39  \n",
       "6         0.42  \n",
       "7         0.52  \n",
       "8         0.75  "
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.read_csv('../results/error-comparison-classification.csv').round(decimals=2).sort_values(by='MCC')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1303,
   "metadata": {},
   "outputs": [],
   "source": [
    "reg_mcc = pd.read_csv('../results/regression-country-comparison-mcc.csv').round(decimals=2).sort_values(by='MCC_10')\n",
    "reg_mcc['Country'] = reg_mcc['name'].str.split('_').str[0]\n",
    "reg_mcc['Model'] = reg_mcc['name'].str.split('_').str[1]\n",
    "reg_mcc['Model'] = reg_mcc['Model'].str.replace('-', ' ').str.capitalize()\n",
    "reg_mcc.rename(columns={'MCC_5': '5 years', 'MCC_10': '10 years', 'MCC_20': '20 years'}, inplace=True)\n",
    "reg_mcc = reg_mcc[['Country', 'Model', '5 years', '10 years', '20 years']]\n",
    "reg_mcc.to_csv('../results/regression-country-comparison-mcc-table.csv', index=False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1304,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "id\n",
      "v0.1-FRA.10.9.2.2_1-1580    11\n",
      "v0.1-FRA.12.2.3.3_1-4597     0\n",
      "v0.1-FRA.9.5.2.9_1-9883     14\n",
      "v0.1-FRA.3.4.1.1_1-6409     12\n",
      "v0.1-FRA.4.4.1.3_1-12521     0\n",
      "                            ..\n",
      "v0.1-FRA.6.7.2.4_1-4027      6\n",
      "v0.1-FRA.12.2.3.5_1-937     22\n",
      "v0.1-FRA.13.6.2.3_1-6862     6\n",
      "v0.1-FRA.1.9.1.6_1-11351    16\n",
      "v0.1-FRA.10.7.4.3_1-6717    13\n",
      "Length: 9596614, dtype: int8\n",
      "id\n",
      "v0.1-FRA.10.9.2.2_1-1580    15\n",
      "v0.1-FRA.12.2.3.3_1-4597     3\n",
      "v0.1-FRA.9.5.2.9_1-9883     16\n",
      "v0.1-FRA.3.4.1.1_1-6409     14\n",
      "v0.1-FRA.4.4.1.3_1-12521     6\n",
      "                            ..\n",
      "v0.1-FRA.6.7.2.4_1-4027      8\n",
      "v0.1-FRA.12.2.3.5_1-937     18\n",
      "v0.1-FRA.13.6.2.3_1-6862     9\n",
      "v0.1-FRA.1.9.1.6_1-11351    13\n",
      "v0.1-FRA.10.7.4.3_1-6717    11\n",
      "Length: 9596614, dtype: int8\n",
      "9596614\n",
      "9596614\n",
      "MCC for FRA with bin size 5: 0.058326614273458054\n",
      "id\n",
      "v0.1-FRA.10.9.2.2_1-1580     5\n",
      "v0.1-FRA.12.2.3.3_1-4597     0\n",
      "v0.1-FRA.9.5.2.9_1-9883      7\n",
      "v0.1-FRA.3.4.1.1_1-6409      6\n",
      "v0.1-FRA.4.4.1.3_1-12521     0\n",
      "                            ..\n",
      "v0.1-FRA.6.7.2.4_1-4027      3\n",
      "v0.1-FRA.12.2.3.5_1-937     11\n",
      "v0.1-FRA.13.6.2.3_1-6862     3\n",
      "v0.1-FRA.1.9.1.6_1-11351     8\n",
      "v0.1-FRA.10.7.4.3_1-6717     6\n",
      "Length: 9596614, dtype: int8\n",
      "id\n",
      "v0.1-FRA.10.9.2.2_1-1580    7\n",
      "v0.1-FRA.12.2.3.3_1-4597    1\n",
      "v0.1-FRA.9.5.2.9_1-9883     8\n",
      "v0.1-FRA.3.4.1.1_1-6409     7\n",
      "v0.1-FRA.4.4.1.3_1-12521    3\n",
      "                           ..\n",
      "v0.1-FRA.6.7.2.4_1-4027     4\n",
      "v0.1-FRA.12.2.3.5_1-937     9\n",
      "v0.1-FRA.13.6.2.3_1-6862    4\n",
      "v0.1-FRA.1.9.1.6_1-11351    6\n",
      "v0.1-FRA.10.7.4.3_1-6717    5\n",
      "Length: 9596614, dtype: int8\n",
      "9596614\n",
      "9596614\n",
      "MCC for FRA with bin size 10: 0.11924236958151267\n",
      "id\n",
      "v0.1-FRA.10.9.2.2_1-1580    2\n",
      "v0.1-FRA.12.2.3.3_1-4597    0\n",
      "v0.1-FRA.9.5.2.9_1-9883     3\n",
      "v0.1-FRA.3.4.1.1_1-6409     3\n",
      "v0.1-FRA.4.4.1.3_1-12521    0\n",
      "                           ..\n",
      "v0.1-FRA.6.7.2.4_1-4027     1\n",
      "v0.1-FRA.12.2.3.5_1-937     5\n",
      "v0.1-FRA.13.6.2.3_1-6862    1\n",
      "v0.1-FRA.1.9.1.6_1-11351    4\n",
      "v0.1-FRA.10.7.4.3_1-6717    3\n",
      "Length: 9596614, dtype: int8\n",
      "id\n",
      "v0.1-FRA.10.9.2.2_1-1580    3\n",
      "v0.1-FRA.12.2.3.3_1-4597    0\n",
      "v0.1-FRA.9.5.2.9_1-9883     4\n",
      "v0.1-FRA.3.4.1.1_1-6409     3\n",
      "v0.1-FRA.4.4.1.3_1-12521    1\n",
      "                           ..\n",
      "v0.1-FRA.6.7.2.4_1-4027     2\n",
      "v0.1-FRA.12.2.3.5_1-937     4\n",
      "v0.1-FRA.13.6.2.3_1-6862    2\n",
      "v0.1-FRA.1.9.1.6_1-11351    3\n",
      "v0.1-FRA.10.7.4.3_1-6717    2\n",
      "Length: 9596614, dtype: int8\n",
      "9596614\n",
      "9596614\n",
      "MCC for FRA with bin size 20: 0.20875894806942133\n",
      "id\n",
      "v0.1-ESP.1.3.4.5_1-46604     21\n",
      "v0.1-ESP.10.3.8.9_1-17785     4\n",
      "v0.1-ESP.5.3.1.19_1-2538      2\n",
      "v0.1-ESP.4.2.3.10_1-313      23\n",
      "v0.1-ESP.12.2.2.11_1-1573     8\n",
      "                             ..\n",
      "v0.1-ESP.4.4.4.10_1-866      21\n",
      "v0.1-ESP.1.7.4.18_1-58377    14\n",
      "v0.1-ESP.1.2.6.4_1-2165       0\n",
      "v0.1-ESP.1.3.2.21_1-6228      8\n",
      "v0.1-ESP.18.1.4.1_1-22209    17\n",
      "Length: 7758976, dtype: int8\n",
      "id\n",
      "v0.1-ESP.1.3.4.5_1-46604     19\n",
      "v0.1-ESP.10.3.8.9_1-17785    12\n",
      "v0.1-ESP.5.3.1.19_1-2538      8\n",
      "v0.1-ESP.4.2.3.10_1-313      15\n",
      "v0.1-ESP.12.2.2.11_1-1573    16\n",
      "                             ..\n",
      "v0.1-ESP.4.4.4.10_1-866      21\n",
      "v0.1-ESP.1.7.4.18_1-58377    17\n",
      "v0.1-ESP.1.2.6.4_1-2165      15\n",
      "v0.1-ESP.1.3.2.21_1-6228     13\n",
      "v0.1-ESP.18.1.4.1_1-22209    16\n",
      "Length: 7758976, dtype: int8\n",
      "7758976\n",
      "7758976\n",
      "MCC for ESP with bin size 5: 0.04074003814851427\n",
      "id\n",
      "v0.1-ESP.1.3.4.5_1-46604     10\n",
      "v0.1-ESP.10.3.8.9_1-17785     2\n",
      "v0.1-ESP.5.3.1.19_1-2538      1\n",
      "v0.1-ESP.4.2.3.10_1-313      11\n",
      "v0.1-ESP.12.2.2.11_1-1573     4\n",
      "                             ..\n",
      "v0.1-ESP.4.4.4.10_1-866      10\n",
      "v0.1-ESP.1.7.4.18_1-58377     7\n",
      "v0.1-ESP.1.2.6.4_1-2165       0\n",
      "v0.1-ESP.1.3.2.21_1-6228      4\n",
      "v0.1-ESP.18.1.4.1_1-22209     8\n",
      "Length: 7758976, dtype: int8\n",
      "id\n",
      "v0.1-ESP.1.3.4.5_1-46604      9\n",
      "v0.1-ESP.10.3.8.9_1-17785     6\n",
      "v0.1-ESP.5.3.1.19_1-2538      4\n",
      "v0.1-ESP.4.2.3.10_1-313       7\n",
      "v0.1-ESP.12.2.2.11_1-1573     8\n",
      "                             ..\n",
      "v0.1-ESP.4.4.4.10_1-866      10\n",
      "v0.1-ESP.1.7.4.18_1-58377     8\n",
      "v0.1-ESP.1.2.6.4_1-2165       7\n",
      "v0.1-ESP.1.3.2.21_1-6228      6\n",
      "v0.1-ESP.18.1.4.1_1-22209     8\n",
      "Length: 7758976, dtype: int8\n",
      "7758976\n",
      "7758976\n",
      "MCC for ESP with bin size 10: 0.08686428681607637\n",
      "id\n",
      "v0.1-ESP.1.3.4.5_1-46604     5\n",
      "v0.1-ESP.10.3.8.9_1-17785    1\n",
      "v0.1-ESP.5.3.1.19_1-2538     0\n",
      "v0.1-ESP.4.2.3.10_1-313      5\n",
      "v0.1-ESP.12.2.2.11_1-1573    2\n",
      "                            ..\n",
      "v0.1-ESP.4.4.4.10_1-866      5\n",
      "v0.1-ESP.1.7.4.18_1-58377    3\n",
      "v0.1-ESP.1.2.6.4_1-2165      0\n",
      "v0.1-ESP.1.3.2.21_1-6228     2\n",
      "v0.1-ESP.18.1.4.1_1-22209    4\n",
      "Length: 7758976, dtype: int8\n",
      "id\n",
      "v0.1-ESP.1.3.4.5_1-46604     4\n",
      "v0.1-ESP.10.3.8.9_1-17785    3\n",
      "v0.1-ESP.5.3.1.19_1-2538     2\n",
      "v0.1-ESP.4.2.3.10_1-313      3\n",
      "v0.1-ESP.12.2.2.11_1-1573    4\n",
      "                            ..\n",
      "v0.1-ESP.4.4.4.10_1-866      5\n",
      "v0.1-ESP.1.7.4.18_1-58377    4\n",
      "v0.1-ESP.1.2.6.4_1-2165      3\n",
      "v0.1-ESP.1.3.2.21_1-6228     3\n",
      "v0.1-ESP.18.1.4.1_1-22209    4\n",
      "Length: 7758976, dtype: int8\n",
      "7758976\n",
      "7758976\n",
      "MCC for ESP with bin size 20: 0.16261303970009716\n",
      "id\n",
      "v0.1-NLD.8.9_1-14183      10\n",
      "v0.1-NLD.8.44_1-7515      17\n",
      "v0.1-NLD.10.4_1-1410      21\n",
      "v0.1-NLD.8.22_1-12033     17\n",
      "v0.1-NLD.3.21_1-14906     20\n",
      "                          ..\n",
      "v0.1-NLD.10.7_1-988       23\n",
      "v0.1-NLD.4.5_1-32450      15\n",
      "v0.1-NLD.11.27_1-41605     6\n",
      "v0.1-NLD.9.3_1-38258      21\n",
      "v0.1-NLD.5.6_1-4292       12\n",
      "Length: 5441430, dtype: int8\n",
      "id\n",
      "v0.1-NLD.8.9_1-14183      14\n",
      "v0.1-NLD.8.44_1-7515      14\n",
      "v0.1-NLD.10.4_1-1410      18\n",
      "v0.1-NLD.8.22_1-12033     14\n",
      "v0.1-NLD.3.21_1-14906     14\n",
      "                          ..\n",
      "v0.1-NLD.10.7_1-988       11\n",
      "v0.1-NLD.4.5_1-32450      18\n",
      "v0.1-NLD.11.27_1-41605     5\n",
      "v0.1-NLD.9.3_1-38258      18\n",
      "v0.1-NLD.5.6_1-4292       18\n",
      "Length: 5441430, dtype: int8\n",
      "5441430\n",
      "5441430\n",
      "MCC for NLD with bin size 5: 0.09297992461823493\n",
      "id\n",
      "v0.1-NLD.8.9_1-14183       5\n",
      "v0.1-NLD.8.44_1-7515       8\n",
      "v0.1-NLD.10.4_1-1410      10\n",
      "v0.1-NLD.8.22_1-12033      8\n",
      "v0.1-NLD.3.21_1-14906     10\n",
      "                          ..\n",
      "v0.1-NLD.10.7_1-988       11\n",
      "v0.1-NLD.4.5_1-32450       7\n",
      "v0.1-NLD.11.27_1-41605     3\n",
      "v0.1-NLD.9.3_1-38258      10\n",
      "v0.1-NLD.5.6_1-4292        6\n",
      "Length: 5441430, dtype: int8\n",
      "id\n",
      "v0.1-NLD.8.9_1-14183      7\n",
      "v0.1-NLD.8.44_1-7515      7\n",
      "v0.1-NLD.10.4_1-1410      9\n",
      "v0.1-NLD.8.22_1-12033     7\n",
      "v0.1-NLD.3.21_1-14906     7\n",
      "                         ..\n",
      "v0.1-NLD.10.7_1-988       5\n",
      "v0.1-NLD.4.5_1-32450      9\n",
      "v0.1-NLD.11.27_1-41605    2\n",
      "v0.1-NLD.9.3_1-38258      9\n",
      "v0.1-NLD.5.6_1-4292       9\n",
      "Length: 5441430, dtype: int8\n",
      "5441430\n",
      "5441430\n",
      "MCC for NLD with bin size 10: 0.1752016059994939\n",
      "id\n",
      "v0.1-NLD.8.9_1-14183      2\n",
      "v0.1-NLD.8.44_1-7515      4\n",
      "v0.1-NLD.10.4_1-1410      5\n",
      "v0.1-NLD.8.22_1-12033     4\n",
      "v0.1-NLD.3.21_1-14906     5\n",
      "                         ..\n",
      "v0.1-NLD.10.7_1-988       5\n",
      "v0.1-NLD.4.5_1-32450      3\n",
      "v0.1-NLD.11.27_1-41605    1\n",
      "v0.1-NLD.9.3_1-38258      5\n",
      "v0.1-NLD.5.6_1-4292       3\n",
      "Length: 5441430, dtype: int8\n",
      "id\n",
      "v0.1-NLD.8.9_1-14183      3\n",
      "v0.1-NLD.8.44_1-7515      3\n",
      "v0.1-NLD.10.4_1-1410      4\n",
      "v0.1-NLD.8.22_1-12033     3\n",
      "v0.1-NLD.3.21_1-14906     3\n",
      "                         ..\n",
      "v0.1-NLD.10.7_1-988       2\n",
      "v0.1-NLD.4.5_1-32450      4\n",
      "v0.1-NLD.11.27_1-41605    1\n",
      "v0.1-NLD.9.3_1-38258      4\n",
      "v0.1-NLD.5.6_1-4292       4\n",
      "Length: 5441430, dtype: int8\n",
      "5441430\n",
      "5441430\n",
      "MCC for NLD with bin size 20: 0.28203282368444893\n"
     ]
    }
   ],
   "source": [
    "\n",
    "for country, predictor in {'FRA': regressor_fts_importance_fr, 'ESP': regressor_fts_importance_es, 'NLD': regressor_fts_importance_nl}.items():\n",
    "    for bin_size in [5, 10, 20]:\n",
    "        bins = utils.generate_bins((1900, 2020, bin_size))\n",
    "        mcc = predictor.mcc(bins)\n",
    "        print(f'MCC for {country} with bin size {bin_size}: {mcc}')\n",
    "\n",
    "# Results\n",
    "# France,Neighborhood cv,0.06,0.12,0.21\n",
    "# Spain,Neighborhood cv,0.04,0.09,0.16\n",
    "# Netherlands,Neighborhood cv,0.09,0.18,0.28"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\">Netherlands</th>\n",
       "      <th colspan=\"2\" halign=\"left\">France</th>\n",
       "      <th colspan=\"2\" halign=\"left\">Spain</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>MCC</th>\n",
       "      <th>F1</th>\n",
       "      <th>MCC</th>\n",
       "      <th>F1</th>\n",
       "      <th>MCC</th>\n",
       "      <th>F1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Bin size</th>\n",
       "      <th>Model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">5 years</th>\n",
       "      <th>Random cv</th>\n",
       "      <td>0.73</td>\n",
       "      <td>0.70</td>\n",
       "      <td>0.37</td>\n",
       "      <td>0.36</td>\n",
       "      <td>0.36</td>\n",
       "      <td>0.37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block cv</th>\n",
       "      <td>0.44</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.22</td>\n",
       "      <td>0.21</td>\n",
       "      <td>0.20</td>\n",
       "      <td>0.22</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>0.21</td>\n",
       "      <td>0.21</td>\n",
       "      <td>0.17</td>\n",
       "      <td>0.16</td>\n",
       "      <td>0.13</td>\n",
       "      <td>0.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>0.17</td>\n",
       "      <td>0.16</td>\n",
       "      <td>0.15</td>\n",
       "      <td>0.14</td>\n",
       "      <td>0.09</td>\n",
       "      <td>0.10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">10 years</th>\n",
       "      <th>Random cv</th>\n",
       "      <td>0.75</td>\n",
       "      <td>0.73</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.45</td>\n",
       "      <td>0.39</td>\n",
       "      <td>0.44</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block cv</th>\n",
       "      <td>0.52</td>\n",
       "      <td>0.51</td>\n",
       "      <td>0.30</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>0.33</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.26</td>\n",
       "      <td>0.28</td>\n",
       "      <td>0.18</td>\n",
       "      <td>0.23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>0.28</td>\n",
       "      <td>0.29</td>\n",
       "      <td>0.24</td>\n",
       "      <td>0.27</td>\n",
       "      <td>0.14</td>\n",
       "      <td>0.19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">20 years</th>\n",
       "      <th>Random cv</th>\n",
       "      <td>0.76</td>\n",
       "      <td>0.77</td>\n",
       "      <td>0.48</td>\n",
       "      <td>0.56</td>\n",
       "      <td>0.45</td>\n",
       "      <td>0.54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block cv</th>\n",
       "      <td>0.58</td>\n",
       "      <td>0.62</td>\n",
       "      <td>0.39</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>0.44</td>\n",
       "      <td>0.51</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.45</td>\n",
       "      <td>0.27</td>\n",
       "      <td>0.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>0.39</td>\n",
       "      <td>0.46</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.43</td>\n",
       "      <td>0.23</td>\n",
       "      <td>0.34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         Netherlands       France       Spain      \n",
       "                                 MCC    F1    MCC    F1   MCC    F1\n",
       "Bin size Model                                                     \n",
       "5 years  Random cv              0.73  0.70   0.37  0.36  0.36  0.37\n",
       "         Block cv               0.44  0.42   0.22  0.21  0.20  0.22\n",
       "         Neighborhood cv        0.21  0.21   0.17  0.16  0.13  0.14\n",
       "         City cv                0.17  0.16   0.15  0.14  0.09  0.10\n",
       "10 years Random cv              0.75  0.73   0.42  0.45  0.39  0.44\n",
       "         Block cv               0.52  0.51   0.30  0.33  0.26  0.31\n",
       "         Neighborhood cv        0.33  0.33   0.26  0.28  0.18  0.23\n",
       "         City cv                0.28  0.29   0.24  0.27  0.14  0.19\n",
       "20 years Random cv              0.76  0.77   0.48  0.56  0.45  0.54\n",
       "         Block cv               0.58  0.62   0.39  0.49  0.35  0.46\n",
       "         Neighborhood cv        0.44  0.51   0.35  0.45  0.27  0.38\n",
       "         City cv                0.39  0.46   0.33  0.43  0.23  0.34"
      ]
     },
     "execution_count": 147,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg_results_mcc = pd.read_csv('../results/regression-country-comparison-mcc-table.csv')\n",
    "class_results_5 = pd.read_csv('../results/error-comparison-classification-5-year-bins.csv').round(decimals=2).sort_values(by='MCC')\n",
    "class_results_10 = pd.read_csv('../results/error-comparison-classification.csv').round(decimals=2).sort_values(by='MCC')\n",
    "class_results_20 = pd.read_csv('../results/error-comparison-classification-20-year-bins.csv').round(decimals=2).sort_values(by='MCC')\n",
    "class_results_5['Bin size'] = '5 years'\n",
    "class_results_10['Bin size'] = '10 years'\n",
    "class_results_20['Bin size'] = '20 years'\n",
    "class_results = pd.concat([class_results_5, class_results_10, class_results_20], ignore_index=True)\n",
    "class_results\n",
    "class_results['Country'] = class_results['name'].str.split('_').str[0]\n",
    "class_results['Model'] = class_results['name'].str.split('_').str[1]\n",
    "class_results['Model'] = class_results['Model'].str.replace('-', ' ').str.capitalize()\n",
    "class_results['Approach'] = 'class'\n",
    "reg_results_mcc['Approach'] = 'reg'\n",
    "reg_results_mcc.columns.name = 'Bin size'\n",
    "\n",
    "# class_results['Recall <1940'] = class_results[['Recall_1900-1909', 'Recall_1910-1919', 'Recall_1920-1929', 'Recall_1930-1939']].mean(axis=1)\n",
    "# class_results['Recall >2000'] = class_results[['Recall_2000-2009', 'Recall_2010-2019']].mean(axis=1)\n",
    "# class_results[['Country', 'Model', 'MCC', 'Recall <1940', 'Recall >2000']]\n",
    "\n",
    "# pivot_table_reg = reg_results_mcc.set_index(['Country', 'Model', 'Approach']).unstack(level=0).swaplevel(0, 1, axis=1).sort_index(axis=1)\n",
    "# pivot_table_class = class_results.pivot_table(values=['MCC'], index=['Model', 'Approach'], columns=['Country', 'Bin size'], margins=False)\n",
    "# pivot_table_class.columns = pivot_table_class.columns.droplevel(0) # drop \"MCC\" multi-index column\n",
    "# pivot_table = pd.concat([pivot_table_class, pivot_table_reg])\n",
    "\n",
    "# pivot_table.columns.name = None # drop Country cell\n",
    "# new_cols = pivot_table.columns.reindex(['Netherlands', 'France', 'Spain'], level=0)\n",
    "# pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "# new_cols = pivot_table.columns.reindex(['5 years', '10 years', '20 years'], level=1)\n",
    "# pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "# pivot_table = pivot_table.sort_index(level=0)\n",
    "# new_cols = pivot_table.index.reindex(['Random cv', 'Block cv', 'Neighborhood cv', 'City cv'], level=0)\n",
    "# pivot_table = pivot_table.reindex(index=new_cols[0])\n",
    "\n",
    "# pivot_table.to_latex('../figures/error-classification.tex', index=True, float_format='%.2f', column_format='|l|l|r|r|r|r|r|r|r|r|r|', caption='MCC for classification and discretized regression', label='table:classification-results')\n",
    "# pivot_table\n",
    "\n",
    "pivot_table_f1 = class_results.pivot_table(values=['MCC', 'F1'], index=['Bin size', 'Model'], columns=['Country'], margins=False).swaplevel(0, 1, axis=1).sort_index(axis=1)\n",
    "pivot_table_f1.columns = pivot_table_f1.columns.rename(None, level=0) # drop Country cell\n",
    "\n",
    "new_cols = pivot_table_f1.columns.reindex(['Netherlands', 'France', 'Spain'], level=0)\n",
    "pivot_table_f1 = pivot_table_f1.reindex(columns=new_cols[0])\n",
    "new_cols = pivot_table_f1.columns.reindex(['MCC', 'F1'], level=1)\n",
    "pivot_table_f1 = pivot_table_f1.reindex(columns=new_cols[0])\n",
    "new_cols = pivot_table_f1.index.reindex(['5 years', '10 years', '20 years'], level=0)\n",
    "pivot_table_f1 = pivot_table_f1.reindex(index=new_cols[0])\n",
    "new_cols = pivot_table_f1.index.reindex(['Random cv', 'Block cv', 'Neighborhood cv', 'City cv'], level=1)\n",
    "pivot_table_f1 = pivot_table_f1.reindex(index=new_cols[0])\n",
    "pivot_table_f1.to_latex('../figures/error-classification-f1-mcc.tex', index=True, float_format='%.2f', column_format='|l|l|r|r|r|r|r|r|', caption='Classification error', label='table:classification-results-f1-mcc')\n",
    "pivot_table_f1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1318,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using the level keyword in DataFrame and Series aggregations is deprecated and will be removed in a future version. Use groupby instead. df.median(level=1) should use df.groupby(level=1).median().\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Bin size\n",
       "5 years     3.660278\n",
       "10 years    2.211593\n",
       "20 years    1.698427\n",
       "dtype: float64"
      ]
     },
     "execution_count": 1318,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "((pivot_table.xs('class', level=1) / pivot_table.xs('reg', level=1))).mean().mean(axis=0, level=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1330,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "121.014082    2\n",
       "280.555723    2\n",
       "112.728860    2\n",
       "83.199405     2\n",
       "105.341157    1\n",
       "             ..\n",
       "10.079536     1\n",
       "22.731155     1\n",
       "15.699179     1\n",
       "65.526033     1\n",
       "207.892964    1\n",
       "Name: FootprintArea, Length: 176315, dtype: int64"
      ]
     },
     "execution_count": 1330,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# oversample(df_nl_sample, None)\n",
    "# df_nl_sample.select_dtypes(include=['object']).columns.\n",
    "df_nl_sample['FootprintArea'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th colspan=\"3\" halign=\"left\">France</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Netherlands</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Spain</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>&lt;1960</th>\n",
       "      <th>1960-1990</th>\n",
       "      <th>&gt;1990</th>\n",
       "      <th>&lt;1960</th>\n",
       "      <th>1960-1990</th>\n",
       "      <th>&gt;1990</th>\n",
       "      <th>&lt;1960</th>\n",
       "      <th>1960-1990</th>\n",
       "      <th>&gt;1990</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Bin size</th>\n",
       "      <th>Model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">5 years</th>\n",
       "      <th>Block cv</th>\n",
       "      <td>19.3</td>\n",
       "      <td>24.7</td>\n",
       "      <td>22.7</td>\n",
       "      <td>34.7</td>\n",
       "      <td>52.8</td>\n",
       "      <td>43.2</td>\n",
       "      <td>26.3</td>\n",
       "      <td>20.7</td>\n",
       "      <td>21.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>13.1</td>\n",
       "      <td>17.7</td>\n",
       "      <td>17.0</td>\n",
       "      <td>9.7</td>\n",
       "      <td>28.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>7.7</td>\n",
       "      <td>10.8</td>\n",
       "      <td>15.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>14.8</td>\n",
       "      <td>18.7</td>\n",
       "      <td>18.0</td>\n",
       "      <td>15.1</td>\n",
       "      <td>30.3</td>\n",
       "      <td>23.7</td>\n",
       "      <td>11.8</td>\n",
       "      <td>13.8</td>\n",
       "      <td>17.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Random cv</th>\n",
       "      <td>33.8</td>\n",
       "      <td>40.0</td>\n",
       "      <td>37.2</td>\n",
       "      <td>67.7</td>\n",
       "      <td>79.3</td>\n",
       "      <td>69.8</td>\n",
       "      <td>43.8</td>\n",
       "      <td>36.7</td>\n",
       "      <td>37.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">10 years</th>\n",
       "      <th>Block cv</th>\n",
       "      <td>31.2</td>\n",
       "      <td>36.3</td>\n",
       "      <td>37.0</td>\n",
       "      <td>43.8</td>\n",
       "      <td>63.3</td>\n",
       "      <td>53.3</td>\n",
       "      <td>36.7</td>\n",
       "      <td>29.3</td>\n",
       "      <td>33.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>24.2</td>\n",
       "      <td>31.0</td>\n",
       "      <td>32.3</td>\n",
       "      <td>18.5</td>\n",
       "      <td>44.7</td>\n",
       "      <td>35.3</td>\n",
       "      <td>15.2</td>\n",
       "      <td>19.7</td>\n",
       "      <td>28.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>26.7</td>\n",
       "      <td>32.0</td>\n",
       "      <td>33.7</td>\n",
       "      <td>24.7</td>\n",
       "      <td>47.7</td>\n",
       "      <td>39.0</td>\n",
       "      <td>20.3</td>\n",
       "      <td>23.0</td>\n",
       "      <td>30.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Random cv</th>\n",
       "      <td>43.5</td>\n",
       "      <td>47.7</td>\n",
       "      <td>47.7</td>\n",
       "      <td>71.7</td>\n",
       "      <td>81.3</td>\n",
       "      <td>73.3</td>\n",
       "      <td>52.5</td>\n",
       "      <td>42.3</td>\n",
       "      <td>44.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">20 years</th>\n",
       "      <th>Block cv</th>\n",
       "      <td>50.0</td>\n",
       "      <td>48.5</td>\n",
       "      <td>56.0</td>\n",
       "      <td>57.7</td>\n",
       "      <td>70.5</td>\n",
       "      <td>66.0</td>\n",
       "      <td>51.7</td>\n",
       "      <td>39.5</td>\n",
       "      <td>58.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>42.3</td>\n",
       "      <td>45.0</td>\n",
       "      <td>52.0</td>\n",
       "      <td>35.0</td>\n",
       "      <td>58.0</td>\n",
       "      <td>54.0</td>\n",
       "      <td>30.3</td>\n",
       "      <td>31.5</td>\n",
       "      <td>59.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>45.3</td>\n",
       "      <td>45.5</td>\n",
       "      <td>53.0</td>\n",
       "      <td>43.3</td>\n",
       "      <td>62.0</td>\n",
       "      <td>55.0</td>\n",
       "      <td>35.7</td>\n",
       "      <td>35.0</td>\n",
       "      <td>59.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Random cv</th>\n",
       "      <td>59.7</td>\n",
       "      <td>55.5</td>\n",
       "      <td>62.0</td>\n",
       "      <td>78.7</td>\n",
       "      <td>82.5</td>\n",
       "      <td>79.0</td>\n",
       "      <td>63.3</td>\n",
       "      <td>49.0</td>\n",
       "      <td>62.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         France                 Netherlands                  \\\n",
       "                          <1960 1960-1990 >1990       <1960 1960-1990 >1990   \n",
       "Bin size Model                                                                \n",
       "5 years  Block cv          19.3      24.7  22.7        34.7      52.8  43.2   \n",
       "         City cv           13.1      17.7  17.0         9.7      28.0  19.0   \n",
       "         Neighborhood cv   14.8      18.7  18.0        15.1      30.3  23.7   \n",
       "         Random cv         33.8      40.0  37.2        67.7      79.3  69.8   \n",
       "10 years Block cv          31.2      36.3  37.0        43.8      63.3  53.3   \n",
       "         City cv           24.2      31.0  32.3        18.5      44.7  35.3   \n",
       "         Neighborhood cv   26.7      32.0  33.7        24.7      47.7  39.0   \n",
       "         Random cv         43.5      47.7  47.7        71.7      81.3  73.3   \n",
       "20 years Block cv          50.0      48.5  56.0        57.7      70.5  66.0   \n",
       "         City cv           42.3      45.0  52.0        35.0      58.0  54.0   \n",
       "         Neighborhood cv   45.3      45.5  53.0        43.3      62.0  55.0   \n",
       "         Random cv         59.7      55.5  62.0        78.7      82.5  79.0   \n",
       "\n",
       "                         Spain                  \n",
       "                         <1960 1960-1990 >1990  \n",
       "Bin size Model                                  \n",
       "5 years  Block cv         26.3      20.7  21.0  \n",
       "         City cv           7.7      10.8  15.5  \n",
       "         Neighborhood cv  11.8      13.8  17.8  \n",
       "         Random cv        43.8      36.7  37.7  \n",
       "10 years Block cv         36.7      29.3  33.0  \n",
       "         City cv          15.2      19.7  28.7  \n",
       "         Neighborhood cv  20.3      23.0  30.7  \n",
       "         Random cv        52.5      42.3  44.0  \n",
       "20 years Block cv         51.7      39.5  58.0  \n",
       "         City cv          30.3      31.5  59.0  \n",
       "         Neighborhood cv  35.7      35.0  59.0  \n",
       "         Random cv        63.3      49.0  62.0  "
      ]
     },
     "execution_count": 149,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "recall_over_time = class_results.set_index(['Bin size', 'Model', 'Country']).filter(regex='Recall')\n",
    "recall_over_time.columns = (recall_over_time.columns.str.split('-').str[0].str.split('_').str[1].astype(float) + recall_over_time.columns.str.split('-').str[1].astype(float) + 1) /2\n",
    "recall_over_time = recall_over_time.transpose().sort_index(axis=1)\n",
    "recall_grouped = pd.concat([recall_over_time[recall_over_time.index < 1960].mean(), recall_over_time[(recall_over_time.index > 1960) & (recall_over_time.index <= 1990)].mean(), recall_over_time[recall_over_time.index > 1990].mean()], keys=['<1960', '1960-1990', '>1990'], axis=1)\n",
    "recall_grouped = recall_grouped.unstack(level=2).swaplevel(0, 1, axis=1).sort_index(axis=1)\n",
    "new_cols = recall_grouped.columns.reindex(['<1960', '1960-1990', '>1990'], level=1)\n",
    "recall_grouped = recall_grouped.reindex(columns=new_cols[0])\n",
    "new_cols = recall_grouped.index.reindex(['5 years', '10 years', '20 years'], level=0)\n",
    "recall_grouped = recall_grouped.reindex(index=new_cols[0])\n",
    "\n",
    "recall_grouped.columns = recall_grouped.columns.rename(None, level=0) # drop Country cell\n",
    "\n",
    "recall_grouped = recall_grouped.multiply(100).round(1)\n",
    "recall_grouped.to_latex('../figures/classification-recall.tex', index=True, float_format='%.2f', caption='Mean classification recall before 1960, after 1990, and in between', label='table:classification-recall')\n",
    "\n",
    "recall_grouped\n",
    "\n",
    "# # performance difference between periods\n",
    "# recall_grouped.groupby(level=0, axis=1).apply(lambda x: x.droplevel(0, axis=1)['<1960'] - x.droplevel(0, axis=1)[['1960-1990', '>1990']].mean(axis=1))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1218,
   "metadata": {},
   "outputs": [],
   "source": [
    "recall_over_time = class_results.set_index(['Bin size', 'Model', 'Country']).filter(regex='Recall')\n",
    "recall_over_time.columns = (recall_over_time.columns.str.split('-').str[0].str.split('_').str[1].astype(float) + recall_over_time.columns.str.split('-').str[1].astype(float) + 1) /2\n",
    "recall_over_time = recall_over_time.transpose()#.swaplevel(0,1, axis=1)\n",
    "\n",
    "new_cols = recall_over_time.columns.reindex(['5 years', '10 years', '20 years'], level=0)\n",
    "recall_over_time = recall_over_time.reindex(columns=new_cols[0])\n",
    "\n",
    "with visualizations.SubplotManager(nrows=3) as axis:\n",
    "    colors = ['darksalmon', 'plum', 'lightseagreen']\n",
    "    cmap = matplotlib.colors.ListedColormap(colors)\n",
    "\n",
    "    for idx, (bin_size, group) in enumerate(recall_over_time.groupby('Bin size', axis=1, sort=False)):\n",
    "        group.droplevel(0, axis=1).plot(y='Neighborhood cv', kind='line', xlabel=None, ylabel=f'Recall\\n({bin_size})', ylim=(0, 0.65), xlim=(1895, 2025), xticks=[1900, 1920, 1940, 1960, 1980, 2000, 2020], cmap=cmap, legend=False, ax=axis[idx])\n",
    "        axis[idx].spines['right'].set_visible(False)\n",
    "        axis[idx].spines['top'].set_visible(False)\n",
    "        # axis[idx].set_xticks([])\n",
    "\n",
    "    pos1 = axis[0].get_position()\n",
    "    pos2 = axis[1].get_position()\n",
    "    pos3 = axis[2].get_position()\n",
    "    axis[0].set_position(pos2)\n",
    "    axis[1].set_position(pos3)\n",
    "    axis[2].set_position(pos1)\n",
    "\n",
    "    axis[2].legend(loc='lower center', bbox_to_anchor=(0.5, 1), fancybox=True, ncol=3, title=None)\n",
    "    axis[1].set_xlabel('Construction period')\n",
    "    axis[1].set_xticks([1900, 1920, 1940, 1960, 1980, 2000, 2020])\n",
    "\n",
    "plt.savefig('../figures/recall-classification.svg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>Kurtosis</th>\n",
       "      <th>Skew</th>\n",
       "      <th>energy_r2</th>\n",
       "      <th>energy_mape</th>\n",
       "      <th>within_5_years</th>\n",
       "      <th>within_10_years</th>\n",
       "      <th>within_20_years</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>total</td>\n",
       "      <td>-0.092903</td>\n",
       "      <td>25.852499</td>\n",
       "      <td>31.832251</td>\n",
       "      <td>742.721340</td>\n",
       "      <td>2.682522</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.117236</td>\n",
       "      <td>0.231709</td>\n",
       "      <td>0.444437</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>spain</td>\n",
       "      <td>0.076176</td>\n",
       "      <td>22.641857</td>\n",
       "      <td>28.662061</td>\n",
       "      <td>0.220374</td>\n",
       "      <td>-0.583579</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.146819</td>\n",
       "      <td>0.287073</td>\n",
       "      <td>0.526781</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>netherlands</td>\n",
       "      <td>-0.113085</td>\n",
       "      <td>27.921449</td>\n",
       "      <td>33.544307</td>\n",
       "      <td>0.019312</td>\n",
       "      <td>-0.515011</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.104943</td>\n",
       "      <td>0.205547</td>\n",
       "      <td>0.393390</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>france</td>\n",
       "      <td>-0.182905</td>\n",
       "      <td>26.766411</td>\n",
       "      <td>32.727943</td>\n",
       "      <td>1454.484728</td>\n",
       "      <td>5.591008</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.106102</td>\n",
       "      <td>0.212196</td>\n",
       "      <td>0.420457</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Unnamed: 0        R2        MAE       RMSE     Kurtosis      Skew  \\\n",
       "0        total -0.092903  25.852499  31.832251   742.721340  2.682522   \n",
       "1        spain  0.076176  22.641857  28.662061     0.220374 -0.583579   \n",
       "2  netherlands -0.113085  27.921449  33.544307     0.019312 -0.515011   \n",
       "3       france -0.182905  26.766411  32.727943  1454.484728  5.591008   \n",
       "\n",
       "   energy_r2  energy_mape  within_5_years  within_10_years  within_20_years  \n",
       "0        NaN          NaN        0.117236         0.231709         0.444437  \n",
       "1        NaN          NaN        0.146819         0.287073         0.526781  \n",
       "2        NaN          NaN        0.104943         0.205547         0.393390  \n",
       "3        NaN          NaN        0.106102         0.212196         0.420457  "
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.read_csv('../results/error-generalization-cross-country-trained-on-single-country.csv')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>Kurtosis</th>\n",
       "      <th>Skew</th>\n",
       "      <th>energy_r2</th>\n",
       "      <th>energy_mape</th>\n",
       "      <th>within_5_years</th>\n",
       "      <th>within_10_years</th>\n",
       "      <th>within_20_years</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>total</td>\n",
       "      <td>-0.023434</td>\n",
       "      <td>25.340801</td>\n",
       "      <td>31.227857</td>\n",
       "      <td>1041.308745</td>\n",
       "      <td>3.780116</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.121550</td>\n",
       "      <td>0.238760</td>\n",
       "      <td>0.450863</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>france</td>\n",
       "      <td>0.134576</td>\n",
       "      <td>23.833616</td>\n",
       "      <td>29.507719</td>\n",
       "      <td>-0.095433</td>\n",
       "      <td>-0.523080</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.140350</td>\n",
       "      <td>0.270125</td>\n",
       "      <td>0.485209</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>spain</td>\n",
       "      <td>-0.145393</td>\n",
       "      <td>28.582238</td>\n",
       "      <td>34.732670</td>\n",
       "      <td>2597.248004</td>\n",
       "      <td>10.586690</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.093458</td>\n",
       "      <td>0.187647</td>\n",
       "      <td>0.377417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>netherlands</td>\n",
       "      <td>-0.178234</td>\n",
       "      <td>23.376912</td>\n",
       "      <td>28.806080</td>\n",
       "      <td>0.483067</td>\n",
       "      <td>-0.526024</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.128450</td>\n",
       "      <td>0.256328</td>\n",
       "      <td>0.495018</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Unnamed: 0        R2        MAE       RMSE     Kurtosis       Skew  \\\n",
       "0        total -0.023434  25.340801  31.227857  1041.308745   3.780116   \n",
       "1       france  0.134576  23.833616  29.507719    -0.095433  -0.523080   \n",
       "2        spain -0.145393  28.582238  34.732670  2597.248004  10.586690   \n",
       "3  netherlands -0.178234  23.376912  28.806080     0.483067  -0.526024   \n",
       "\n",
       "   energy_r2  energy_mape  within_5_years  within_10_years  within_20_years  \n",
       "0        NaN          NaN        0.121550         0.238760         0.450863  \n",
       "1        NaN          NaN        0.140350         0.270125         0.485209  \n",
       "2        NaN          NaN        0.093458         0.187647         0.377417  \n",
       "3        NaN          NaN        0.128450         0.256328         0.495018  "
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.read_csv('../results/error-generalization-across-countries-jul-5.csv')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "invalid value encountered in true_divide\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1008x720 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Pass the following variables as keyword args: x, y. From version 0.12, the only valid positional argument will be `data`, and passing other arguments without an explicit keyword will result in an error or misinterpretation.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x432 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# gen_pred = pd.read_pickle('../predictions-23818532-20220705-15-09-45-exp.pkl')\n",
    "pred = gen_pred.iloc[:, 0].to_frame('age')\n",
    "test = gen_pred.iloc[:, 1].to_frame('age')\n",
    "bins = list(range(1900, 2025 + 1))[0::5]\n",
    "# gen_pred.columns = ['predict', 'test']\n",
    "\n",
    "_, axis = plt.subplots(2, 2, figsize=(14, 10), constrained_layout=True)\n",
    "visualizations.plot_histogram(test, pred, bins=bins, ax=axis[1, 0])\n",
    "visualizations.plot_relative_grid(test, pred, ax=axis[1, 1])\n",
    "plt.show()\n",
    "visualizations.plot_grid(test, pred)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "      <th>support</th>\n",
       "      <th>kappa</th>\n",
       "      <th>mcc</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>energy_r2</th>\n",
       "      <th>energy_mape</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.336218</td>\n",
       "      <td>0.618282</td>\n",
       "      <td>0.435574</td>\n",
       "      <td>865473.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.451313</td>\n",
       "      <td>0.440838</td>\n",
       "      <td>0.446014</td>\n",
       "      <td>1003939.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.423944</td>\n",
       "      <td>0.421660</td>\n",
       "      <td>0.422799</td>\n",
       "      <td>1311321.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.292099</td>\n",
       "      <td>0.245631</td>\n",
       "      <td>0.266857</td>\n",
       "      <td>967631.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.334375</td>\n",
       "      <td>0.295591</td>\n",
       "      <td>0.313789</td>\n",
       "      <td>1206461.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.313218</td>\n",
       "      <td>0.248659</td>\n",
       "      <td>0.277230</td>\n",
       "      <td>1118813.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.275284</td>\n",
       "      <td>0.224501</td>\n",
       "      <td>0.247313</td>\n",
       "      <td>1039290.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.258885</td>\n",
       "      <td>0.229602</td>\n",
       "      <td>0.243366</td>\n",
       "      <td>797825.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.316537</td>\n",
       "      <td>0.267379</td>\n",
       "      <td>0.289889</td>\n",
       "      <td>954765.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.184614</td>\n",
       "      <td>0.329034</td>\n",
       "      <td>0.236521</td>\n",
       "      <td>331096.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.318649</td>\n",
       "      <td>0.332118</td>\n",
       "      <td>0.317935</td>\n",
       "      <td>9596614.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.332667</td>\n",
       "      <td>0.331767</td>\n",
       "      <td>0.326408</td>\n",
       "      <td>9596614.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.254449</td>\n",
       "      <td>0.255623</td>\n",
       "      <td>0.331767</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.255908</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.257681</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.254354</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.255143</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.254968</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    precision    recall  f1-score    support     kappa       mcc  accuracy  \\\n",
       "0    0.336218  0.618282  0.435574   865473.0       NaN       NaN       NaN   \n",
       "1    0.451313  0.440838  0.446014  1003939.0       NaN       NaN       NaN   \n",
       "2    0.423944  0.421660  0.422799  1311321.0       NaN       NaN       NaN   \n",
       "3    0.292099  0.245631  0.266857   967631.0       NaN       NaN       NaN   \n",
       "4    0.334375  0.295591  0.313789  1206461.0       NaN       NaN       NaN   \n",
       "5    0.313218  0.248659  0.277230  1118813.0       NaN       NaN       NaN   \n",
       "6    0.275284  0.224501  0.247313  1039290.0       NaN       NaN       NaN   \n",
       "7    0.258885  0.229602  0.243366   797825.0       NaN       NaN       NaN   \n",
       "8    0.316537  0.267379  0.289889   954765.0       NaN       NaN       NaN   \n",
       "9    0.184614  0.329034  0.236521   331096.0       NaN       NaN       NaN   \n",
       "10   0.318649  0.332118  0.317935  9596614.0       NaN       NaN       NaN   \n",
       "11   0.332667  0.331767  0.326408  9596614.0       NaN       NaN       NaN   \n",
       "12        NaN       NaN       NaN        NaN  0.254449  0.255623  0.331767   \n",
       "13        NaN       NaN       NaN        NaN       NaN  0.255908       NaN   \n",
       "14        NaN       NaN       NaN        NaN       NaN  0.257681       NaN   \n",
       "15        NaN       NaN       NaN        NaN       NaN  0.254354       NaN   \n",
       "16        NaN       NaN       NaN        NaN       NaN  0.255143       NaN   \n",
       "17        NaN       NaN       NaN        NaN       NaN  0.254968       NaN   \n",
       "\n",
       "    energy_r2  energy_mape  \n",
       "0         NaN          NaN  \n",
       "1         NaN          NaN  \n",
       "2         NaN          NaN  \n",
       "3         NaN          NaN  \n",
       "4         NaN          NaN  \n",
       "5         NaN          NaN  \n",
       "6         NaN          NaN  \n",
       "7         NaN          NaN  \n",
       "8         NaN          NaN  \n",
       "9         NaN          NaN  \n",
       "10        NaN          NaN  \n",
       "11        NaN          NaN  \n",
       "12        NaN          NaN  \n",
       "13        NaN          NaN  \n",
       "14        NaN          NaN  \n",
       "15        NaN          NaN  \n",
       "16        NaN          NaN  \n",
       "17        NaN          NaN  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.read_csv('../error-class-energy-error-23811418-20220704-02-26-47.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>MCC</th>\n",
       "      <th>F1</th>\n",
       "      <th>Recall_1900-1904</th>\n",
       "      <th>Recall_1905-1909</th>\n",
       "      <th>Recall_1910-1914</th>\n",
       "      <th>Recall_1915-1919</th>\n",
       "      <th>Recall_1920-1924</th>\n",
       "      <th>Recall_1925-1929</th>\n",
       "      <th>Recall_1930-1934</th>\n",
       "      <th>...</th>\n",
       "      <th>Recall_1970-1974</th>\n",
       "      <th>Recall_1975-1979</th>\n",
       "      <th>Recall_1980-1984</th>\n",
       "      <th>Recall_1985-1989</th>\n",
       "      <th>Recall_1990-1994</th>\n",
       "      <th>Recall_1995-1999</th>\n",
       "      <th>Recall_2000-2004</th>\n",
       "      <th>Recall_2005-2009</th>\n",
       "      <th>Recall_2010-2014</th>\n",
       "      <th>Recall_2015-2019</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Spain_city-cv</td>\n",
       "      <td>0.079885</td>\n",
       "      <td>0.098834</td>\n",
       "      <td>0.325625</td>\n",
       "      <td>0.160035</td>\n",
       "      <td>0.066501</td>\n",
       "      <td>0.118466</td>\n",
       "      <td>0.151644</td>\n",
       "      <td>0.089388</td>\n",
       "      <td>0.123557</td>\n",
       "      <td>...</td>\n",
       "      <td>0.058642</td>\n",
       "      <td>0.124095</td>\n",
       "      <td>0.063851</td>\n",
       "      <td>0.093923</td>\n",
       "      <td>0.054136</td>\n",
       "      <td>0.069356</td>\n",
       "      <td>0.077834</td>\n",
       "      <td>0.167138</td>\n",
       "      <td>0.207791</td>\n",
       "      <td>0.310153</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>France_city-cv</td>\n",
       "      <td>0.122474</td>\n",
       "      <td>0.141126</td>\n",
       "      <td>0.382910</td>\n",
       "      <td>0.146016</td>\n",
       "      <td>0.107226</td>\n",
       "      <td>0.050009</td>\n",
       "      <td>0.278541</td>\n",
       "      <td>0.115447</td>\n",
       "      <td>0.115364</td>\n",
       "      <td>...</td>\n",
       "      <td>0.132085</td>\n",
       "      <td>0.193401</td>\n",
       "      <td>0.128001</td>\n",
       "      <td>0.157136</td>\n",
       "      <td>0.092925</td>\n",
       "      <td>0.136392</td>\n",
       "      <td>0.132738</td>\n",
       "      <td>0.162461</td>\n",
       "      <td>0.212439</td>\n",
       "      <td>0.298990</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Spain_random-cv</td>\n",
       "      <td>0.157520</td>\n",
       "      <td>0.180753</td>\n",
       "      <td>0.469290</td>\n",
       "      <td>0.460736</td>\n",
       "      <td>0.321576</td>\n",
       "      <td>0.437423</td>\n",
       "      <td>0.347301</td>\n",
       "      <td>0.339065</td>\n",
       "      <td>0.275482</td>\n",
       "      <td>...</td>\n",
       "      <td>0.107495</td>\n",
       "      <td>0.177722</td>\n",
       "      <td>0.102893</td>\n",
       "      <td>0.157204</td>\n",
       "      <td>0.100409</td>\n",
       "      <td>0.101137</td>\n",
       "      <td>0.102948</td>\n",
       "      <td>0.196309</td>\n",
       "      <td>0.246137</td>\n",
       "      <td>0.399632</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>France_random-cv</td>\n",
       "      <td>0.161607</td>\n",
       "      <td>0.182643</td>\n",
       "      <td>0.407974</td>\n",
       "      <td>0.224531</td>\n",
       "      <td>0.174827</td>\n",
       "      <td>0.158829</td>\n",
       "      <td>0.333214</td>\n",
       "      <td>0.177674</td>\n",
       "      <td>0.164008</td>\n",
       "      <td>...</td>\n",
       "      <td>0.170551</td>\n",
       "      <td>0.226703</td>\n",
       "      <td>0.155427</td>\n",
       "      <td>0.191672</td>\n",
       "      <td>0.123815</td>\n",
       "      <td>0.166537</td>\n",
       "      <td>0.153895</td>\n",
       "      <td>0.174960</td>\n",
       "      <td>0.228058</td>\n",
       "      <td>0.334042</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Netherlands_city-cv</td>\n",
       "      <td>0.161633</td>\n",
       "      <td>0.165653</td>\n",
       "      <td>0.282331</td>\n",
       "      <td>0.171944</td>\n",
       "      <td>0.162001</td>\n",
       "      <td>0.068045</td>\n",
       "      <td>0.062405</td>\n",
       "      <td>0.110113</td>\n",
       "      <td>0.092211</td>\n",
       "      <td>...</td>\n",
       "      <td>0.303125</td>\n",
       "      <td>0.261436</td>\n",
       "      <td>0.228702</td>\n",
       "      <td>0.194522</td>\n",
       "      <td>0.168878</td>\n",
       "      <td>0.187392</td>\n",
       "      <td>0.150508</td>\n",
       "      <td>0.151107</td>\n",
       "      <td>0.219330</td>\n",
       "      <td>0.337367</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Netherlands_random-cv</td>\n",
       "      <td>0.414989</td>\n",
       "      <td>0.400669</td>\n",
       "      <td>0.542977</td>\n",
       "      <td>0.491024</td>\n",
       "      <td>0.385125</td>\n",
       "      <td>0.480158</td>\n",
       "      <td>0.333041</td>\n",
       "      <td>0.386558</td>\n",
       "      <td>0.344099</td>\n",
       "      <td>...</td>\n",
       "      <td>0.483511</td>\n",
       "      <td>0.464229</td>\n",
       "      <td>0.494593</td>\n",
       "      <td>0.428917</td>\n",
       "      <td>0.438422</td>\n",
       "      <td>0.455592</td>\n",
       "      <td>0.416631</td>\n",
       "      <td>0.398452</td>\n",
       "      <td>0.384535</td>\n",
       "      <td>0.612028</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>6 rows × 27 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                    name       MCC        F1  Recall_1900-1904  \\\n",
       "0          Spain_city-cv  0.079885  0.098834          0.325625   \n",
       "1         France_city-cv  0.122474  0.141126          0.382910   \n",
       "2        Spain_random-cv  0.157520  0.180753          0.469290   \n",
       "3       France_random-cv  0.161607  0.182643          0.407974   \n",
       "4    Netherlands_city-cv  0.161633  0.165653          0.282331   \n",
       "5  Netherlands_random-cv  0.414989  0.400669          0.542977   \n",
       "\n",
       "   Recall_1905-1909  Recall_1910-1914  Recall_1915-1919  Recall_1920-1924  \\\n",
       "0          0.160035          0.066501          0.118466          0.151644   \n",
       "1          0.146016          0.107226          0.050009          0.278541   \n",
       "2          0.460736          0.321576          0.437423          0.347301   \n",
       "3          0.224531          0.174827          0.158829          0.333214   \n",
       "4          0.171944          0.162001          0.068045          0.062405   \n",
       "5          0.491024          0.385125          0.480158          0.333041   \n",
       "\n",
       "   Recall_1925-1929  Recall_1930-1934  ...  Recall_1970-1974  \\\n",
       "0          0.089388          0.123557  ...          0.058642   \n",
       "1          0.115447          0.115364  ...          0.132085   \n",
       "2          0.339065          0.275482  ...          0.107495   \n",
       "3          0.177674          0.164008  ...          0.170551   \n",
       "4          0.110113          0.092211  ...          0.303125   \n",
       "5          0.386558          0.344099  ...          0.483511   \n",
       "\n",
       "   Recall_1975-1979  Recall_1980-1984  Recall_1985-1989  Recall_1990-1994  \\\n",
       "0          0.124095          0.063851          0.093923          0.054136   \n",
       "1          0.193401          0.128001          0.157136          0.092925   \n",
       "2          0.177722          0.102893          0.157204          0.100409   \n",
       "3          0.226703          0.155427          0.191672          0.123815   \n",
       "4          0.261436          0.228702          0.194522          0.168878   \n",
       "5          0.464229          0.494593          0.428917          0.438422   \n",
       "\n",
       "   Recall_1995-1999  Recall_2000-2004  Recall_2005-2009  Recall_2010-2014  \\\n",
       "0          0.069356          0.077834          0.167138          0.207791   \n",
       "1          0.136392          0.132738          0.162461          0.212439   \n",
       "2          0.101137          0.102948          0.196309          0.246137   \n",
       "3          0.166537          0.153895          0.174960          0.228058   \n",
       "4          0.187392          0.150508          0.151107          0.219330   \n",
       "5          0.455592          0.416631          0.398452          0.384535   \n",
       "\n",
       "   Recall_2015-2019  \n",
       "0          0.310153  \n",
       "1          0.298990  \n",
       "2          0.399632  \n",
       "3          0.334042  \n",
       "4          0.337367  \n",
       "5          0.612028  \n",
       "\n",
       "[6 rows x 27 columns]"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.read_csv('../results/results-classification-23785961-20220622-05-38-45.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Spain_city-cv</td>\n",
       "      <td>0.305322</td>\n",
       "      <td>20.297378</td>\n",
       "      <td>27.362372</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Spain_random-cv</td>\n",
       "      <td>0.339042</td>\n",
       "      <td>19.682322</td>\n",
       "      <td>26.690025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Netherlands_city-cv</td>\n",
       "      <td>0.344058</td>\n",
       "      <td>15.875018</td>\n",
       "      <td>21.444882</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>France_city-cv</td>\n",
       "      <td>0.371999</td>\n",
       "      <td>19.101783</td>\n",
       "      <td>25.396953</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>France_random-cv</td>\n",
       "      <td>0.385177</td>\n",
       "      <td>18.827284</td>\n",
       "      <td>25.129069</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Netherlands_random-cv</td>\n",
       "      <td>0.451107</td>\n",
       "      <td>14.061322</td>\n",
       "      <td>19.617100</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    name        R2        MAE       RMSE\n",
       "0          Spain_city-cv  0.305322  20.297378  27.362372\n",
       "1        Spain_random-cv  0.339042  19.682322  26.690025\n",
       "2    Netherlands_city-cv  0.344058  15.875018  21.444882\n",
       "3         France_city-cv  0.371999  19.101783  25.396953\n",
       "4       France_random-cv  0.385177  18.827284  25.129069\n",
       "5  Netherlands_random-cv  0.451107  14.061322  19.617100"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.read_csv('../results/results-regression-23785966-20220622-01-48-08.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 257,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "['name'] did not aggregate successfully. If any error is raised this will raise in a future version of pandas. Drop these columns/ops to avoid this warning.\n",
      "['name'] did not aggregate successfully. If any error is raised this will raise in a future version of pandas. Drop these columns/ops to avoid this warning.\n"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [],
   "source": [
    "regressor_sd = AgePredictor.load('../model-spatial-distance-23839686-20220709-04-14-17-exp_125-Flevoland__0_.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>174893.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>1988.629139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>13.466541</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1916.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>1981.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>1989.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>1998.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>2016.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 age\n",
       "count  174893.000000\n",
       "mean     1988.629139\n",
       "std        13.466541\n",
       "min      1916.000000\n",
       "25%      1981.000000\n",
       "50%      1989.000000\n",
       "75%      1998.000000\n",
       "max      2016.000000"
      ]
     },
     "execution_count": 141,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat([regressor_sd.y_predict, regressor_sd.y_test], axis=1).sort_index()\n",
    "regressor_sd.y_test.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Strategy</th>\n",
       "      <th>MCC</th>\n",
       "      <th>F1</th>\n",
       "      <th>Recall &lt;1960</th>\n",
       "      <th>Recall &gt;1990</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>undersampling</td>\n",
       "      <td>0.653</td>\n",
       "      <td>0.614</td>\n",
       "      <td>0.591500</td>\n",
       "      <td>0.663000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>oversampling</td>\n",
       "      <td>0.714</td>\n",
       "      <td>0.664</td>\n",
       "      <td>0.609667</td>\n",
       "      <td>0.719000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>SMOTE</td>\n",
       "      <td>0.715</td>\n",
       "      <td>0.665</td>\n",
       "      <td>0.582333</td>\n",
       "      <td>0.726000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ADASYN</td>\n",
       "      <td>0.715</td>\n",
       "      <td>0.663</td>\n",
       "      <td>0.576500</td>\n",
       "      <td>0.733667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>baseline</td>\n",
       "      <td>0.715</td>\n",
       "      <td>0.667</td>\n",
       "      <td>0.547667</td>\n",
       "      <td>0.726000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>sample weights</td>\n",
       "      <td>0.716</td>\n",
       "      <td>0.667</td>\n",
       "      <td>0.615500</td>\n",
       "      <td>0.722333</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         Strategy    MCC     F1  Recall <1960  Recall >1990\n",
       "0   undersampling  0.653  0.614      0.591500      0.663000\n",
       "1    oversampling  0.714  0.664      0.609667      0.719000\n",
       "2           SMOTE  0.715  0.665      0.582333      0.726000\n",
       "3          ADASYN  0.715  0.663      0.576500      0.733667\n",
       "5        baseline  0.715  0.667      0.547667      0.726000\n",
       "4  sample weights  0.716  0.667      0.615500      0.722333"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "resample_results = pd.read_csv('../results/prelim-comparison-resampling-strategies.csv').sort_values(by='MCC').round(3)\n",
    "resample_results['Country'] = resample_results['name'].str.split('_').str[1]\n",
    "resample_results['Strategy'] = resample_results['name'].str.split('_').str[0]\n",
    "resample_results['Recall <1960'] = resample_results[['Recall_1900-1909', 'Recall_1910-1919', 'Recall_1920-1929', 'Recall_1930-1939', 'Recall_1940-1949', 'Recall_1950-1959']].mean(axis=1)\n",
    "resample_results['Recall >1990'] = resample_results[['Recall_1990-1999', 'Recall_2000-2009', 'Recall_2010-2019']].mean(axis=1)\n",
    "resample_results[['Strategy', 'MCC', 'F1', 'Recall <1960', 'Recall >1990']].to_latex('../figures/resampling-strategies.tex', index=False, float_format='%.3f', caption='Comparison of resampling strategies for imbalanced classification', label='table:resampling')\n",
    "resample_results[['Strategy', 'MCC', 'F1', 'Recall <1960', 'Recall >1990']]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Model selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Model</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "      <th>Fit time [s]</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>AdaBoost</td>\n",
       "      <td>18.512</td>\n",
       "      <td>24.001</td>\n",
       "      <td>0.211</td>\n",
       "      <td>409</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>RandomForest</td>\n",
       "      <td>9.160</td>\n",
       "      <td>17.039</td>\n",
       "      <td>0.602</td>\n",
       "      <td>722</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>XGboost</td>\n",
       "      <td>8.925</td>\n",
       "      <td>16.657</td>\n",
       "      <td>0.620</td>\n",
       "      <td>281</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Model     MAE    RMSE     R2 Fit time [s]\n",
       "0      AdaBoost  18.512  24.001  0.211          409\n",
       "2  RandomForest   9.160  17.039  0.602          722\n",
       "1       XGboost   8.925  16.657  0.620          281"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_selection = pd.read_csv('../results/model-selection.csv').sort_values(by='R2').round(3)\n",
    "model_selection['Model'] = model_selection['name'].str.split('_').str[1]\n",
    "model_selection.loc[model_selection['Model'] == 'RandomForest', 'Fit time [s]'] = '722'\n",
    "model_selection.loc[model_selection['Model'] == 'XGboost', 'Fit time [s]'] = '281'\n",
    "model_selection.loc[model_selection['Model'] == 'AdaBoost', 'Fit time [s]'] = '409'\n",
    "model_selection = model_selection[['Model', 'MAE', 'RMSE', 'R2', 'Fit time [s]']]\n",
    "model_selection.to_latex('../figures/model-selection.tex', index=False, float_format='%.3f', caption='Model comparison between Random Forest, XGBoost, and AdaBoost for regression', label='table:model-selection')\n",
    "model_selection\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Figures"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure: Descriptive statistics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 683,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>share</th>\n",
       "      <th>settlement size</th>\n",
       "      <th>mean age</th>\n",
       "      <th>std age</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>country</th>\n",
       "      <th>DEGURBA</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">ESP</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>23.352812</td>\n",
       "      <td>14117.019608</td>\n",
       "      <td>1978.569354</td>\n",
       "      <td>27.007953</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>39.330234</td>\n",
       "      <td>3819.062992</td>\n",
       "      <td>1981.385893</td>\n",
       "      <td>28.820408</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>37.316954</td>\n",
       "      <td>655.172551</td>\n",
       "      <td>1968.157417</td>\n",
       "      <td>37.455753</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>non-residential</th>\n",
       "      <td>21.646957</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1979.015622</td>\n",
       "      <td>32.380960</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>residential</th>\n",
       "      <td>78.353043</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1974.621136</td>\n",
       "      <td>32.264286</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">FRA</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>7.975371</td>\n",
       "      <td>5433.177419</td>\n",
       "      <td>1959.053926</td>\n",
       "      <td>32.108184</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>17.751276</td>\n",
       "      <td>4760.403175</td>\n",
       "      <td>1971.476161</td>\n",
       "      <td>29.627559</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>74.273353</td>\n",
       "      <td>3019.341193</td>\n",
       "      <td>1973.742303</td>\n",
       "      <td>31.871257</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>non-residential</th>\n",
       "      <td>17.920877</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1966.543449</td>\n",
       "      <td>34.304319</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>residential</th>\n",
       "      <td>82.079123</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1973.144001</td>\n",
       "      <td>31.001249</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">NLD</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>43.054494</td>\n",
       "      <td>23800.326531</td>\n",
       "      <td>1970.476185</td>\n",
       "      <td>27.921076</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>41.574724</td>\n",
       "      <td>9463.306723</td>\n",
       "      <td>1974.429724</td>\n",
       "      <td>24.577314</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>15.370782</td>\n",
       "      <td>9462.454545</td>\n",
       "      <td>1973.223723</td>\n",
       "      <td>27.151679</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>non-residential</th>\n",
       "      <td>100.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1972.504306</td>\n",
       "      <td>26.541734</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                              share  settlement size     mean age    std age\n",
       "country DEGURBA                                                             \n",
       "ESP     1 (city)          23.352812     14117.019608  1978.569354  27.007953\n",
       "        2 (town)          39.330234      3819.062992  1981.385893  28.820408\n",
       "        3 (rural)         37.316954       655.172551  1968.157417  37.455753\n",
       "        non-residential   21.646957              NaN  1979.015622  32.380960\n",
       "        residential       78.353043              NaN  1974.621136  32.264286\n",
       "FRA     1 (city)           7.975371      5433.177419  1959.053926  32.108184\n",
       "        2 (town)          17.751276      4760.403175  1971.476161  29.627559\n",
       "        3 (rural)         74.273353      3019.341193  1973.742303  31.871257\n",
       "        non-residential   17.920877              NaN  1966.543449  34.304319\n",
       "        residential       82.079123              NaN  1973.144001  31.001249\n",
       "NLD     1 (city)          43.054494     23800.326531  1970.476185  27.921076\n",
       "        2 (town)          41.574724      9463.306723  1974.429724  24.577314\n",
       "        3 (rural)         15.370782      9462.454545  1973.223723  27.151679\n",
       "        non-residential  100.000000              NaN  1972.504306  26.541734"
      ]
     },
     "execution_count": 683,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# predictors = {'Netherlands': regressor_fts_importance_nl, 'Spain': regressor_fts_importance_es, 'France': regressor_fts_importance_fr, 'All': regressor_country_cv}\n",
    "\n",
    "# for country, predictor in predictors.items():\n",
    "#     print(predictor.y_test.describe())\n",
    "\n",
    "# df_countries = []\n",
    "# for predictor, gadm, gadm_level in [(regressor_fts_importance_fr, gadm_france, 4), (regressor_fts_importance_es, gadm_spain, 4), (regressor_fts_importance_nl, gadm_netherlands, 2)]:\n",
    "#     df_country = pd.concat([predictor.y_test, predictor.aux_vars_test, predictor.X_test[['FootprintArea']]], axis=1)\n",
    "#     df_country = df_country.merge(gadm[[f'NAME_{gadm_level}', 'DEGURBA']], right_on=f'NAME_{gadm_level}', left_on='city', how='left').set_index('id')\n",
    "#     df_countries.append(df_country)\n",
    "\n",
    "# data = pd.concat(df_countries, axis=0)\n",
    "# data = harmonize_group_source_types(data)\n",
    "# data['residential'] = data['type'] == 'residential'\n",
    "# data.replace({1.0: '1 (city)', 2.0: '2 (town)', 3.0: '3 (rural)'}, inplace=True)\n",
    "\n",
    "# data = pd.concat([regressor_country_cv.y_test, regressor_country_cv.aux_vars_test], axis=1)\n",
    "# gadm = pd.concat([gadm_netherlands, gadm_spain, gadm_france], axis=0)\n",
    "# # gadm\n",
    "# bins=[1900, 1944, 1990, 2022]\n",
    "# # data.groupby('country')['age'].apply(['mean', 'std', 'count', lambda x: np.histogram(x, bins=bins)[0] / len(x)])\n",
    "# data.groupby('country')['age'].agg(['mean', 'std', 'count', lambda x: x.quantile(0.25), lambda x: x.quantile(0.75)])\n",
    "# data.groupby(['country', 'DEGURBA'])['age'].agg(['mean', 'std', 'count', lambda x: x.quantile(0.25), lambda x: x.quantile(0.75)])\n",
    "# data.groupby(['country', 'residential'])['age'].agg(['mean', 'std', 'count', lambda x: x.quantile(0.25), lambda x: x.quantile(0.75)])\n",
    "# data.groupby('country')['age'].value_counts(bins=bins, normalize=True, sort=False)\n",
    "\n",
    "blds_per_degurba = data.groupby(['country', 'DEGURBA', 'city'])['age'].count().reset_index().groupby(['country', 'DEGURBA']).mean().rename(columns={'age' :'settlement size'})\n",
    "\n",
    "age_per_degurba = data.groupby(['country', 'DEGURBA'])['age'].agg(['mean', 'std']).rename(columns={'mean': 'mean age', 'std': 'std age'})\n",
    "share_per_degurba = data.groupby(['country'])['DEGURBA'].value_counts(normalize=True).rename('share').multiply(100)\n",
    "age_per_type = data.groupby(['country', 'residential'])['age'].agg(['mean', 'std']).rename(index={True: 'residential', False: 'non-residential'}, level=1).rename(columns={'mean': 'mean age', 'std': 'std age'})\n",
    "share_per_type = data.groupby(['country'])['residential'].value_counts(normalize=True).rename('share').multiply(100).rename(index={True: 'residential', False: 'non-residential'}, level=1)\n",
    "type_table = pd.concat([age_per_type, share_per_type], axis=1).sort_index()\n",
    "degurba_table = pd.concat([share_per_degurba, blds_per_degurba, age_per_degurba], axis=1).sort_index()\n",
    "table = pd.concat([degurba_table, type_table], axis=0).sort_index()\n",
    "table.rename(index={1.0: '1 (city)', 2.0: '2 (town)', 3.0: '3 (rural)'}, level=1, inplace=True)\n",
    "\n",
    "table.to_latex('../figures/detailed-stats.tex', float_format='%.1f', index=True, multicolumn_format='c', column_format='|l|r|rr|rr|', caption='Settlement and building type statistics of preprocessed dataset', label='table:detailed-stats')\n",
    "\n",
    "table\n",
    "# regressor_fts_importance_nl.X_test\n",
    "# regressor_fts_importance_nl.y_test.hist(bins=[1900, 1944, 1990, 2022])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 676,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\"></th>\n",
       "      <th colspan=\"4\" halign=\"left\">age</th>\n",
       "      <th colspan=\"4\" halign=\"left\">other properties</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>country</th>\n",
       "      <th>n</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>Q1</th>\n",
       "      <th>Q3</th>\n",
       "      <th>residential [%]</th>\n",
       "      <th>footprint area [m2]</th>\n",
       "      <th>height [m]</th>\n",
       "      <th>DEGURBA</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ESP</td>\n",
       "      <td>7994380.0</td>\n",
       "      <td>1975.6</td>\n",
       "      <td>32.3</td>\n",
       "      <td>1960.0</td>\n",
       "      <td>2001.0</td>\n",
       "      <td>78.4</td>\n",
       "      <td>184.6</td>\n",
       "      <td>5.1</td>\n",
       "      <td>2.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>FRA</td>\n",
       "      <td>9644500.0</td>\n",
       "      <td>1972.0</td>\n",
       "      <td>31.7</td>\n",
       "      <td>1958.0</td>\n",
       "      <td>1997.0</td>\n",
       "      <td>82.1</td>\n",
       "      <td>131.2</td>\n",
       "      <td>5.0</td>\n",
       "      <td>2.66</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>NLD</td>\n",
       "      <td>5452189.0</td>\n",
       "      <td>1972.5</td>\n",
       "      <td>26.5</td>\n",
       "      <td>1960.0</td>\n",
       "      <td>1992.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>122.9</td>\n",
       "      <td>5.6</td>\n",
       "      <td>1.72</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         age                       other properties  \\\n",
       "  country          n    mean   std      Q1      Q3  residential [%]   \n",
       "0     ESP  7994380.0  1975.6  32.3  1960.0  2001.0             78.4   \n",
       "1     FRA  9644500.0  1972.0  31.7  1958.0  1997.0             82.1   \n",
       "2     NLD  5452189.0  1972.5  26.5  1960.0  1992.0              0.0   \n",
       "\n",
       "                                          \n",
       "  footprint area [m2] height [m] DEGURBA  \n",
       "0               184.6        5.1    2.14  \n",
       "1               131.2        5.0    2.66  \n",
       "2               122.9        5.6    1.72  "
      ]
     },
     "execution_count": 676,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "overview_table = data.groupby(['country'])[['age', 'FootprintArea', 'DEGURBA', 'height', 'residential', 'city']].apply(lambda x: pd.Series({\n",
    "    'n': int(x['age'].size),\n",
    "    'mean': x['age'].mean().round(1),\n",
    "    'std': x['age'].std().round(1),\n",
    "    'Q1': int(x['age'].quantile(0.25)),\n",
    "    'Q3': int(x['age'].quantile(0.75)),\n",
    "    'residential [%]': ((x['residential'].sum() / x['residential'].size) * 100).round(1),\n",
    "    'footprint area [m2]': x['FootprintArea'].mean().round(1),\n",
    "    'height [m]': x['height'].mean().round(1),\n",
    "    # 'DEGURBA': x['DEGURBA'].mean().round(2),\n",
    "    # **x.groupby(['DEGURBA'])['age'].mean().round(1).to_dict(),\n",
    "    # **x.groupby(['DEGURBA', 'city'])['age'].count().groupby(['DEGURBA']).mean().astype(int).to_dict(),\n",
    "\n",
    "})).reset_index()\n",
    "\n",
    "# overview_table.rename(columns={1.0: 'city size', 2.0: 'town size', 3.0: 'rural size'}, inplace=True)\n",
    "# overview_table.rename(columns={1.0: 'city mean age', 2.0: 'town size', 3.0: 'rural size'}, inplace=True)\n",
    "new_level = ['', '', 'Construction year', 'Construction year', 'Construction year', 'Construction year', 'Other properties', 'Other properties', 'Other properties']\n",
    "overview_table.columns = pd.MultiIndex.from_tuples(zip(new_level, overview_table.columns))\n",
    "\n",
    "# overview_table.to_latex('../figures/stats.tex', index=False, multicolumn_format='c', column_format='|l|r|rrrr|rrrr|', caption='Summary statistics of preprocessed dataset', label='table:stats')\n",
    "overview_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 677,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultiIndex([(                '',             'country'),\n",
       "            (                '',                   'n'),\n",
       "            (             'age',                'mean'),\n",
       "            (             'age',                 'std'),\n",
       "            (             'age',                  'Q1'),\n",
       "            (             'age',                  'Q3'),\n",
       "            ('other properties',     'residential [%]'),\n",
       "            ('other properties', 'footprint area [m2]'),\n",
       "            ('other properties',          'height [m]'),\n",
       "            ('other properties',             'DEGURBA')],\n",
       "           )"
      ]
     },
     "execution_count": 677,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "overview_table.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 665,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"2\" halign=\"left\"></th>\n",
       "      <th colspan=\"4\" halign=\"left\">age</th>\n",
       "      <th colspan=\"4\" halign=\"left\"></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>(nan, country)</th>\n",
       "      <th>(nan, n)</th>\n",
       "      <th>(age, mean)</th>\n",
       "      <th>(age, std)</th>\n",
       "      <th>(age, Q1)</th>\n",
       "      <th>(age, Q3)</th>\n",
       "      <th>(nan, residential [%])</th>\n",
       "      <th>(nan, footprint area [m2])</th>\n",
       "      <th>(nan, height [m])</th>\n",
       "      <th>(nan, DEGURBA)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ESP</td>\n",
       "      <td>7994380.0</td>\n",
       "      <td>1975.6</td>\n",
       "      <td>32.3</td>\n",
       "      <td>1960.0</td>\n",
       "      <td>2001.0</td>\n",
       "      <td>78.4</td>\n",
       "      <td>184.6</td>\n",
       "      <td>5.1</td>\n",
       "      <td>2.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>FRA</td>\n",
       "      <td>9644500.0</td>\n",
       "      <td>1972.0</td>\n",
       "      <td>31.7</td>\n",
       "      <td>1958.0</td>\n",
       "      <td>1997.0</td>\n",
       "      <td>82.1</td>\n",
       "      <td>131.2</td>\n",
       "      <td>5.0</td>\n",
       "      <td>2.66</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>NLD</td>\n",
       "      <td>5452189.0</td>\n",
       "      <td>1972.5</td>\n",
       "      <td>26.5</td>\n",
       "      <td>1960.0</td>\n",
       "      <td>1992.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>122.9</td>\n",
       "      <td>5.6</td>\n",
       "      <td>1.72</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                    age                                 \\\n",
       "  (nan, country)   (nan, n) (age, mean) (age, std) (age, Q1) (age, Q3)   \n",
       "0            ESP  7994380.0      1975.6       32.3    1960.0    2001.0   \n",
       "1            FRA  9644500.0      1972.0       31.7    1958.0    1997.0   \n",
       "2            NLD  5452189.0      1972.5       26.5    1960.0    1992.0   \n",
       "\n",
       "                                                                       \\\n",
       "  (nan, residential [%]) (nan, footprint area [m2]) (nan, height [m])   \n",
       "0                   78.4                      184.6               5.1   \n",
       "1                   82.1                      131.2               5.0   \n",
       "2                    0.0                      122.9               5.6   \n",
       "\n",
       "                  \n",
       "  (nan, DEGURBA)  \n",
       "0           2.14  \n",
       "1           2.66  \n",
       "2           1.72  "
      ]
     },
     "execution_count": 665,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "overview_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 649,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{1.0: 13722.475308641975, 2.0: 5199.409932659933, 3.0: 1656.8955618175414}"
      ]
     },
     "execution_count": 649,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.groupby(['DEGURBA', 'city'])['age'].count().groupby(['DEGURBA']).mean().to_dict()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "82.64437689969604\n",
      "88.7981711299804\n",
      "74.84172859895402\n"
     ]
    }
   ],
   "source": [
    "\n",
    "with open(os.path.join('..', 'metadata', f'buildings-per-region-FRA.json'), 'r', encoding='utf-8') as f:\n",
    "    buildings_per_city = json.load(f)\n",
    "\n",
    "# buildings per neighborhood\n",
    "# buildings per city\n",
    "# buildings per urban / street block\n",
    "\n",
    "setA = set(gadm_france['NAME_4'])\n",
    "setB = set( buildings_per_city.keys())\n",
    "\n",
    "overlap = setA & setB\n",
    "universe = setA | setB\n",
    "\n",
    "print(float(len(overlap)) / len(setA) * 100)\n",
    "print(float(len(overlap)) / len(setB) * 100)\n",
    "print(float(len(overlap)) / len(universe) * 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 246,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_nl_full = pd.read_pickle('../data/exp/df-NLD.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_nl_full_info = df_nl_full.merge(gadm_netherlands[['NAME_2', 'DEGURBA']], left_on='city', right_on='NAME_2', how='left')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0    2717175\n",
       "2.0    2547584\n",
       "3.0     965954\n",
       "Name: DEGURBA, dtype: int64"
      ]
     },
     "execution_count": 183,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_nl_full_info['DEGURBA'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 414,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "mean    1973.349803\n",
       "std       30.846128\n",
       "Name: age, dtype: float64"
      ]
     },
     "execution_count": 414,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat(dfs)['age'].agg(['mean', 'std'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "del df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                   block_areas: 92.0 MiB\n",
      "                    gadm_spain: 37.6 MiB\n",
      "                            _9: 23.9 MiB\n",
      "                   gadm_france: 16.8 MiB\n",
      "                            _8: 12.9 MiB\n",
      "            neighborhood_areas: 12.9 MiB\n",
      "                           _29: 12.9 MiB\n",
      "                           _30: 12.9 MiB\n",
      "              gadm_netherlands:  3.3 MiB\n",
      "               city_boundaries:  3.3 MiB\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "def sizeof_fmt(num, suffix='B'):\n",
    "    ''' by Fred Cirera,  https://stackoverflow.com/a/1094933/1870254, modified'''\n",
    "    for unit in ['','Ki','Mi','Gi','Ti','Pi','Ei','Zi']:\n",
    "        if abs(num) < 1024.0:\n",
    "            return \"%3.1f %s%s\" % (num, unit, suffix)\n",
    "        num /= 1024.0\n",
    "    return \"%.1f %s%s\" % (num, 'Yi', suffix)\n",
    "\n",
    "for name, size in sorted(((name, sys.getsizeof(value)) for name, value in locals().items()),\n",
    "                         key= lambda x: -x[1])[:10]:\n",
    "    print(\"{:>30}: {:>8}\".format(name, sizeof_fmt(size)))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib qt\n",
    "import matplotlib\n",
    "\n",
    "# stats = []\n",
    "# dfs = []\n",
    "# for predictor, city_boundaries, gadm_level, dissolve_level in [(regressor_fts_importance_fr, gadm_france, 4, 2), (regressor_fts_importance_es, gadm_spain, 4, 2), (regressor_fts_importance_nl, gadm_netherlands, 2, 1)]:\n",
    "#     df = pd.concat([predictor.y_test, predictor.aux_vars_test], axis=1, join=\"inner\")\n",
    "#     df = df.merge(city_boundaries[[f'NAME_{gadm_level}', 'DEGURBA', f'NAME_{dissolve_level}']], right_on=f'NAME_{gadm_level}', left_on='city', how='left').reset_index()\n",
    "\n",
    "#     region_stats = df.groupby([f'NAME_{dissolve_level}'], dropna=False).apply(lambda x: pd.Series({\n",
    "#         'age_mean'       : x['age'].mean(),\n",
    "#         'age_std'        : x['age'].std(),\n",
    "#         'DEGURBA_mean'   : x['DEGURBA'].mean(),\n",
    "#         'n'            : len(x.index),\n",
    "#     })).reset_index()\n",
    "\n",
    "#     region_stats = geometry.to_gdf(city_boundaries).dissolve(by=f'GID_{dissolve_level}').merge(region_stats, on=f'NAME_{dissolve_level}', how='left')\n",
    "#     stats.append(region_stats)\n",
    "#     dfs.append(df)\n",
    "\n",
    "# eu_stats = pd.concat(stats)\n",
    "\n",
    "# colors = ['plum', 'thistle', 'lavenderblush', 'paleturquoise', 'mediumturquoise', 'lightseagreen']\n",
    "# colors = ['plum', 'thistle', 'lavenderblush', 'paleturquoise', 'mediumturquoise']\n",
    "colors = ['plum', 'thistle', 'lavenderblush']\n",
    "cmap = matplotlib.colors.LinearSegmentedColormap.from_list(\"\", colors)\n",
    "country_colors = ['darksalmon', 'steelblue', 'teal']\n",
    "\n",
    "# AGE MAP\n",
    "# with visualizations.SubplotManager() as ax1:\n",
    "#     eu_stats.dissolve('GID_0').boundary.plot(ax=ax1, color='thistle', linewidth=0.8)\n",
    "\n",
    "#     cmap_boundaries = matplotlib.colors.ListedColormap(country_colors)\n",
    "#     gpd.GeoDataFrame(geometry=eu_stats.dissolve('GID_0').boundary.simplify(2500)).reset_index().plot(ax=ax1, column='GID_0', cmap=cmap_boundaries, linewidth=0.5)\n",
    "    \n",
    "#     eu_stats = eu_stats.fillna(eu_stats['age_mean'].mean())\n",
    "#     eu_stats.plot(ax=ax1, column='age_mean', legend=True, cmap=cmap, edgecolor=None, linewidth=0,  legend_kwds={'shrink': 0.7, 'pad': 0, 'label': 'Construction year'})\n",
    "#     ax1.set_axis_off()\n",
    "\n",
    "\n",
    "for i, c in enumerate(['France', 'Spain', 'Netherlands']):\n",
    "    sns.set_theme(palette=colors, style='white', font_scale=4)\n",
    "\n",
    "\n",
    "#     # AGE HISTOGRAM\n",
    "    with visualizations.SubplotManager() as ax:\n",
    "        colors = ['steelblue', 'darksalmon', 'teal']\n",
    "        df = dfs[i]\n",
    "        df = df[df[dataset.AGE_ATTRIBUTE] < 2020]\n",
    "        sns.histplot(x='age', data=df, ax=ax, color=colors[i], legend=False, stat='density', binwidth=5)\n",
    "        ax.set_xlabel(f'Construction year {c}')\n",
    "        ax.set_xticks([1900, 1940, 1980, 2020], [1900, 1940, 1980, 2020])\n",
    "        ax.spines['top'].set_visible(False)\n",
    "        ax.spines['right'].set_visible(False)\n",
    "        ax.spines['left'].set_linewidth(False)\n",
    "        ax.spines['bottom'].set_linewidth(0.5)\n",
    "        ax.get_yaxis().set_visible(False)\n",
    "        plt.savefig(f'../figures/age-hist-{c}.svg', bbox_inches='tight')\n",
    "\n",
    "\n",
    "#     # DEGURBA PIE CHART\n",
    "    # with visualizations.SubplotManager() as ax:\n",
    "    #     # colors = ['lavender', 'thistle', 'lightsteelblue']\n",
    "    #     country_cmap = matplotlib.colors.LinearSegmentedColormap.from_list(\"\", ['white', country_colors[i]])\n",
    "    #     pie_colors = [country_cmap(j) for j in [0.3, 0.6, 0.9]]\n",
    "    #     city_types = dfs[i]['DEGURBA'].dropna().value_counts(normalize=True, sort=False).to_dict()\n",
    "    #     ax.pie(city_types.values(), labels=['rural', 'town', 'city'], explode=(0, 0, 0.1), colors=pie_colors, autopct='%1.1f%%', startangle=90, textprops={'fontsize': 48})\n",
    "    #     # ax.set_title(f'Distribution of settlement types in {c} (DEGURBA)', y=0, pad=-30, verticalalignment='top', fontsize=45)\n",
    "    #     ax.set_axis_off()\n",
    "    #     plt.savefig(f'../figures/DEGURBA-pie-{c}.svg')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Spatial cross-validation illustration\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 520,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from shapely.geometry import MultiPoint, LineString\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "\n",
    "#     gpd.GeoSeries(LineString([(0.5, 0.5), (1, 1)]), crs=3035).plot(ax=ax, color='violet', linestyle='dashed', alpha=0.6)\n",
    "    buffer.total_bounds\n",
    "    minx, miny, maxx, maxy = buffer.total_bounds\n",
    "    y = (miny + maxy) / 2\n",
    "    gpd.GeoSeries(LineString([[minx, y], [maxx, y]]), crs=3035).plot(ax=ax, color='violet', linestyle='dashed', alpha=0.6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 527,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-29 13:48:12,058 | WARNING : Dataset is already a GeoDataFrame. Skipping convertion.\n"
     ]
    }
   ],
   "source": [
    "from shapely.geometry import LineString\n",
    "\n",
    "predictor = regressor_fts_importance_nl\n",
    "y = predictor.y_test.rename(columns={'age': 'y_test'})\n",
    "y = pd.concat([y, predictor.aux_vars_test], axis=1, join=\"inner\")\n",
    "y = y[y['city'] == 'Ede'] # Rennes, Limoges, Maastricht, Almelo\n",
    "y['id'] = y['id_source']\n",
    "y = y.drop(columns=['geometry'])\n",
    "\n",
    "y_geom = geometry.add_geometry_column(y, crs=3035, countries=['netherlands'])\n",
    "y_geom = geometry.to_gdf(y_geom)\n",
    "dataset.GLOBAL_REPRODUCIBILITY_SEED = 2\n",
    "\n",
    "\n",
    "cleaned = y_geom.groupby('neighborhood').filter(lambda x: len(x) > 100)\n",
    "train, test = split_by_neighborhood(cleaned)\n",
    "\n",
    "%matplotlib qt\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    # cleaned.dissolve('neighborhood').convex_hull.plot(ax=ax, color='thistle', alpha=0.3) # optional\n",
    "    train.plot(ax=ax, column='neighborhood', cmap='Purples', legend=False, edgecolor=None)\n",
    "    test.plot(ax=ax, column='neighborhood', cmap='Greens', legend=False, edgecolor=None)\n",
    "\n",
    "    for n in ['524679ef', 'da5bd9bd', '5f6df977']:\n",
    "        buffer = test[test['neighborhood'] == n].dissolve()['geometry'].centroid.buffer(900)\n",
    "        buffer.boundary.plot(ax=ax, color='violet', linewidth=2, alpha=0.6)\n",
    "        minx, miny, maxx, maxy = buffer.total_bounds\n",
    "        y1 = miny + (maxy - miny) / 6.5\n",
    "        y2 = maxy - (maxy - miny) / 6.5\n",
    "        x1 = minx + (maxx - minx) / 6.5\n",
    "        x2 = maxx - (maxx - minx) / 6.5\n",
    "        # gpd.GeoSeries(LineString([[minx, y], [maxx, y]]), crs=3035).plot(ax=ax, color='violet', linestyle='dashed', alpha=0.6)\n",
    "        gpd.GeoSeries(LineString([[x1, y1], [x2, y2]]), crs=3035).plot(ax=ax, color='violet', linestyle='dashed', alpha=0.6)\n",
    "\n",
    "    plt.axis('off')\n",
    "    from matplotlib.lines import Line2D\n",
    "    import matplotlib.patches as mpatches\n",
    "    diameter_patch = Line2D([0], [0], color='violet', linewidth=1.5, linestyle='dashed', label='1km neighborhood diameter')\n",
    "    train_patch = mpatches.Patch(color='purple', label='Train set')\n",
    "    test_patch = mpatches.Patch(color='green', label='Test set')\n",
    "    ax.legend(loc='upper center', fancybox=True, handles=[train_patch, test_patch, diameter_patch], fontsize=12)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.7294772544722075"
      ]
     },
     "execution_count": 70,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cleaned.dissolve('neighborhood').convex_hull.area.mean() / 1000**2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.785"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "3.14 * 0.5 ** 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Neighorhood / block size analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "take up more space, the distance between city centroids is X.X times larger than in France and Spain.\n",
    "\n",
    "distacne matrix for regions and consider 4 nearest neighbors and average their centroid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14.59118488838008\n",
      "6.740008159806675\n",
      "178.86468843661243\n",
      "26.540103496650037\n"
     ]
    }
   ],
   "source": [
    "neighborhoods_per_city = regressor_country_cv.aux_vars_test.groupby(['city'])['neighborhood'].agg('nunique').mean()\n",
    "blocks_per_neighborhood = regressor_country_cv.aux_vars_test.groupby(['neighborhood'])['sbb'].agg('nunique').mean()\n",
    "\n",
    "buildings_per_neighborhood = regressor_country_cv.aux_vars_test['neighborhood'].value_counts().mean()\n",
    "buildings_per_block = regressor_country_cv.aux_vars_test['sbb'].value_counts().mean()\n",
    "\n",
    "print(neighborhoods_per_city)\n",
    "print(blocks_per_neighborhood)\n",
    "print(buildings_per_neighborhood)\n",
    "print(buildings_per_block)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.0518372237201\n",
      "0.3085103782234442\n",
      "6.650788331775407\n",
      "3724649.2734119077\n",
      "4410704.633672597\n"
     ]
    }
   ],
   "source": [
    "# aux_vars_geo = geometry.lat_lon_to_gdf(regressor_country_cv.aux_vars_test[['lat', 'lon','neighborhood', 'sbb', 'city', 'country']])\n",
    "# aux_vars_geo = geometry.to_gdf(regressor_country_cv.aux_vars_test[['geometry', 'neighborhood', 'sbb', 'city', 'country']])\n",
    "# neighborhood_areas = aux_vars_geo.dissolve('neighborhood').convex_hull.area\n",
    "# block_areas = aux_vars_geo.dissolve('sbb').convex_hull.area\n",
    "\n",
    "avg_neighborhood_size = neighborhood_areas[neighborhood_areas > 0].mean() / 1000**2\n",
    "avg_block_size = block_areas[block_areas > 0].mean() / 1000**2\n",
    "\n",
    "print(avg_neighborhood_size)\n",
    "print(avg_block_size)\n",
    "print(avg_neighborhood_size / avg_block_size)\n",
    "\n",
    "print(neighborhood_areas[neighborhood_areas > 0].std())\n",
    "print(block_areas[block_areas > 0].std())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "60.95443985978995"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# geometry.to_gdf(gadm_netherlands).dissolve('GID_2').area.mean() / 1000**2\n",
    "# geometry.to_gdf(gadm_france).dissolve('GID_4').area.mean() / 1000**2\n",
    "geometry.to_gdf(gadm_spain).dissolve('GID_4').area.mean() / 1000**2\n",
    "# groupby('GID_1').agg({'GID_1': 'nunique'}).mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure: Regression vs. Classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import metrics\n",
    "import preprocessing\n",
    "\n",
    "def classification_report_for_regression(y_test, y_predict, bins=[], bin_config=None):\n",
    "    bins = bins or utils.generate_bins(bin_config)\n",
    "    labels = utils.generate_labels(bins)\n",
    "    print(bins)\n",
    "    print(labels)\n",
    "    y_test_cat = preprocessing.categorize_age(y_test, bins=bins, remove_outliers=True)\n",
    "    y_predict_cat = preprocessing.categorize_age(y_predict.loc[y_test_cat.index], bins=bins, remove_outliers=False)\n",
    "    # labels.insert(0, 'Outlier')\n",
    "    print(metrics.classification_report(y_test_cat, y_predict_cat[[dataset.AGE_ATTRIBUTE]], target_names=labels))\n",
    "    print(metrics.matthews_corrcoef(y_test_cat, y_predict_cat[[dataset.AGE_ATTRIBUTE]]))\n",
    "    print(metrics.cohen_kappa_score(y_test_cat, y_predict_cat[[dataset.AGE_ATTRIBUTE]]))\n",
    "    visualizations.plot_histogram(y_test_cat, y_predict_cat[[dataset.AGE_ATTRIBUTE]], bins=list(range(0, len(bins))), bin_labels=labels)\n",
    "    visualizations.plot_confusion_matrix(y_test_cat, y_predict_cat[[dataset.AGE_ATTRIBUTE]], class_labels=labels)\n",
    "    plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sklearn.metrics as metrics\n",
    "def mean_bin_age(s, bins):\n",
    "    cat_to_metric = {idx: (bins[idx] + bins[idx+1]) / 2 for idx in range(len(bins) - 1)}\n",
    "    return s.map(cat_to_metric)\n",
    "\n",
    "def rmse(y_test, y_predict):\n",
    "    return np.sqrt(metrics.mean_squared_error(y_test, y_predict))\n",
    "\n",
    "def mae(y_test, y_predict):\n",
    "    return metrics.mean_absolute_error(y_test, y_predict)\n",
    "\n",
    "def r2(y_test, y_predict):\n",
    "    return metrics.r2_score(y_test, y_predict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "regressor_nl = AgePredictor.load('../results/model-reg-nl-neighborhood-23850997-20220708-11-04-41-exp.pkl')\n",
    "classifier_nl = AgeClassifier.load('../results/model-class-nl-neighborhood-23855379-20220710-04-16-58-exp.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25.98501095535683\n",
      "15.983773566874884\n",
      "0.04124170987008813\n"
     ]
    }
   ],
   "source": [
    "bins = utils.generate_bins(bin_config=(1900, 2020, 10))\n",
    "for classifier in [classifier_nl]:\n",
    "    y_test_metric = classifier.aux_vars_test[classifier.metric_target_attribute]\n",
    "    y_predict_metric = mean_bin_age(classifier.y_predict[classifier.target_attribute], classifier.bins)\n",
    "    print(rmse(y_test_metric, y_predict_metric))\n",
    "    print(mae(y_test_metric, y_predict_metric))\n",
    "    print(r2(y_test_metric, y_predict_metric))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "# classification_report_for_regression(regressor_fts_importance_fr.y_test.copy(), regressor_fts_importance_fr.y_predict.copy(), bins=[0, 1910, 1920, 1930, 1940, 1950, 1960, 1970, 1980, 1990, 2000, 2010, 2050])\n",
    "# classification_report_for_regression(regressor_fts_importance_es.y_test.copy(), regressor_fts_importance_es.y_predict.copy(), bins=[0, 1910, 1920, 1930, 1940, 1950, 1960, 1970, 1980, 1990, 2000, 2010, 2050])\n",
    "# classification_report_for_regression(regressor_fts_importance_nl.y_test.copy(), regressor_fts_importance_nl.y_predict.copy(), bins=[0, 1910, 1920, 1930, 1940, 1950, 1960, 1970, 1980, 1990, 2000, 2010, 2050])\n",
    "# classification_report_for_regression(regressor_nl.y_test.copy(), regressor_nl.y_predict.copy(), bins=[0, 1910, 1920, 1930, 1940, 1950, 1960, 1970, 1980, 1990, 2000, 2010, 2050])\n",
    "# plt.savefig('../figures/confusion-matrix-for-regression-nl-neighborhood.svg')\n",
    "\n",
    "# classifier_nl.mcc()\n",
    "visualizations.plot_confusion_matrix(classifier_nl.y_test, classifier_nl.y_predict[[classifier_nl.target_attribute]], class_labels=classifier_nl.labels)\n",
    "plt.savefig('../figures/confusion-matrix-nl-neighborhood.svg')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 390,
   "metadata": {},
   "outputs": [],
   "source": [
    "# residential type column missing\n",
    "regressor_fr = AgePredictor.load('../results/model-reg-energy-error-23879948-20220724-01-06-14-exp.pkl')\n",
    "classifier_fr = AgeClassifier.load('../results/model-class-energy-error-23833213-20220707-01-25-29-exp.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 389,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'classifier_fr' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb Cell 79'\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000263?line=0'>1</a>\u001b[0m classifier_fr\u001b[39m.\u001b[39maux_vars_test\n",
      "\u001b[0;31mNameError\u001b[0m: name 'classifier_fr' is not defined"
     ]
    }
   ],
   "source": [
    "classifier_fr.aux_vars_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 386,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['source_file', 'geometry', 'block', 'country', 'height', 'Unnamed: 0.1',\n",
       "       'sbb', 'id_source', 'city', 'id_misaligned', 'id', 'type_source',\n",
       "       'floors', 'TouchesIndexes', 'type', 'neighborhood', 'n_neighbors',\n",
       "       'residential_type', 'cv_fold_idx'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 386,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "regressor_fr.aux_vars_test.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 393,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "path = os.path.join('..', 'results', 'residential_type_dict_FRA.pkl')\n",
    "with open(path, 'rb') as f:\n",
    "    mapping = pickle.load(f)\n",
    "\n",
    "# regressor_fts_importance_fr.aux_vars_test['residential_type'] = regressor_fts_importance_fr.aux_vars_test['id'].map(mapping)\n",
    "# regressor_fr.aux_vars_test['residential_type'] = regressor_fr.aux_vars_test['id'].map(mapping)\n",
    "classifier_fr.aux_vars_test['residential_type'] = classifier_fr.aux_vars_test['id'].map(mapping)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 388,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-25 17:23:42,844 | INFO : R2: 0.0921\n",
      "2022-07-25 17:23:42,903 | INFO : MAPE: 0.5751\n",
      "2022-07-25 17:23:42,938 | INFO : MAE: 63.11\n",
      "2022-07-25 17:23:43,002 | INFO : RMSE: 101.80\n"
     ]
    }
   ],
   "source": [
    "import energy_modeling\n",
    "\n",
    "y_true = pd.concat([regressor_fr.y_test, regressor_fr.aux_vars_test], axis=1, join=\"inner\")\n",
    "y_predict = pd.concat([regressor_fr.y_predict, regressor_fr.aux_vars_test], axis=1, join=\"inner\")\n",
    "\n",
    "y_true = energy_modeling.assign_heating_energy_demand(y_true, labels=None)\n",
    "y_predict = energy_modeling.assign_heating_energy_demand(y_predict, labels=None)\n",
    "\n",
    "ids = y_true.index.intersection(y_predict.index)\n",
    "y_true_new = y_true.loc[ids]\n",
    "y_predict_new = y_predict.loc[ids]\n",
    "\n",
    "len(y_true_new) - len(y_true)\n",
    "\n",
    "\n",
    "r2 = metrics.r2_score(y_true_new['heating_demand'], y_predict_new['heating_demand'])\n",
    "mape = metrics.mean_absolute_percentage_error(y_true_new['heating_demand'], y_predict_new['heating_demand'])\n",
    "mae = metrics.mean_absolute_error(y_true_new['heating_demand'], y_predict_new['heating_demand'])\n",
    "rmse = np.sqrt(metrics.mean_squared_error(y_true_new['heating_demand'], y_predict_new['heating_demand']))\n",
    "\n",
    "logger.info(f'R2: {r2:.4f}')\n",
    "logger.info(f'MAPE: {mape:.4f}')\n",
    "logger.info(f'MAE: {mae:.2f}')\n",
    "logger.info(f'RMSE: {rmse:.2f}')\n",
    "\n",
    "# 2022-07-22 15:42:33,338 | INFO : R2: 0.0112\n",
    "# 2022-07-22 15:42:33,414 | INFO : MAPE: 0.6257\n",
    "# 2022-07-22 15:42:33,418 | INFO : MAE: 67.90\n",
    "# 2022-07-22 15:42:33,421 | INFO : RMSE: 106.24"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 395,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-25 18:33:35,610 | INFO : R2: 0.0271\n",
      "2022-07-25 18:33:35,620 | INFO : MAPE: 0.5559\n",
      "2022-07-25 18:33:35,649 | INFO : MAE: 63.01\n",
      "2022-07-25 18:33:35,681 | INFO : RMSE: 105.38\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(0.027093233297119546, 0.5559165368221187)"
      ]
     },
     "execution_count": 395,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# classification_report_for_regression(regressor_fr.y_test.copy(), regressor_fr.y_predict.copy(), bins=dataset.TABULA_AGE_BINS['france'])\n",
    "# # plt.savefig('../figures/confusion-matrix-for-regression-fr-random-energy.svg')\n",
    "\n",
    "classifier_fr.energy_error()\n",
    "# visualizations.plot_confusion_matrix(classifier_fr.y_test, classifier_fr.y_predict[[classifier_fr.target_attribute]], class_labels=classifier_fr.labels)\n",
    "# plt.savefig('../figures/confusion-matrix-fr-random-energy.svg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 396,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0776067290906206"
      ]
     },
     "execution_count": 396,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "67.90 / 63.01"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure: Impact of additional data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib qt\n",
    "df_test = df#.sample(frac=0.1)\n",
    "df_test = df_test[df_test['y_test'] <= 2020]\n",
    "df_test = df_test[df_test['y_predict'] <= 2030]\n",
    "# # df_test['DEGURBA'].replace({'1 (city)': 1, '2 (town)':2 , '3 (rural)': 3}, inplace=True)\n",
    "\n",
    "# # df_test['error'] = df_test['y_predict'] - df_test['y_test']\n",
    "df_test = df_test.rename(columns={'NAME_0': 'Country'})\n",
    "df_test['urban'] = df_test['DEGURBA'] < 3.0\n",
    "df_test.loc[df_test['type'] == 'non-residential', 'residential_type'] = 'non-residential'\n",
    "\n",
    "# with visualizations.SubplotManager() as ax:\n",
    "#     sns.violinplot(data=df_test[df_test['Country'] == 'France'], x='residential_type', y='error', inner='quartile', split=True, palette='husl', order=['SFH', 'TH', 'MFH', 'AB', 'non-residential'], ax=ax)\n",
    "#     ax.set_xlabel('Building type')\n",
    "#     ax.set_ylabel('Prediction error [years]')\n",
    "    # plt.savefig('../figures/error-by-residential-type-FRA-violin.svg')\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    df_test['DEGURBA'].replace({1.0: '1 (city)', 2.0: '2 (town)', 3.0: '3 (rural)'}, inplace=True)\n",
    "    sns.violinplot(data=df_test[df_test['Country'] == 'France'], x='DEGURBA', y='error', hue='type', inner='quartile', split=True, palette='husl', order=['1 (city)', '2 (town)', '3 (rural)'], ax=ax)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "    plt.savefig('../figures/error-by-settlement-type-FRA-violin.svg')\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    sns.violinplot(data=df_test, x='Country', y='error', hue='urban', inner='quartile', split=True, palette='husl', order=['Netherlands', 'France', 'Spain'], ax=ax)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    sns.violinplot(data=df_test, x='DEGURBA', y='error', hue='Country', inner='quartile', split=False, palette='husl', ax=ax)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    sns.violinplot(data=df_test[df_test['type'].isin(['residential', 'non-residential'])], x='Country', y='error', hue='type', inner='quartile', split=True, palette='husl', order=['Netherlands', 'France', 'Spain'], ax=ax)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    df_test['type'].fillna('unknown', inplace=True)\n",
    "    order = ['residential', 'non-residential', 'unknown']\n",
    "    sns.violinplot(data=df_test[df_test['Country'].isin(['France', 'Spain'])], x='type', y='error', hue='Country', order=order, inner='quartile', split=True, palette='husl', ax=ax)\n",
    "    ax.set_xlabel('Building type', labelpad=20)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "    ax.axvline(0.5, color='grey', linestyle='--', lw=0.75)\n",
    "    ax.axvline(1.5, color='grey', linestyle='--', lw=0.75)\n",
    "    plt.title('Prediction error for building types in France and Spain')\n",
    "#     plt.savefig('../figures/error-by-type-country-violin.svg')\n",
    "\n",
    "\n",
    "\n",
    "# with visualizations.SubplotManager(figsize=(19, 9)) as ax: # 16,9 for Spain\n",
    "#     order = ['SFH', 'TH', 'MFH', 'AB', 'commercial', 'agricultural', 'industrial', 'others', 'unknown']\n",
    "\n",
    "#     country, annot_pos = 'France', [0.22, 0.67]\n",
    "#     # country, annot_pos = 'Spain', [0.25, 0.69]\n",
    "#     # order.remove('others')\n",
    "\n",
    "#     df_country = df_test[df_test['Country'] == country]\n",
    "\n",
    "#     df_country = harmonize_group_source_types(df_country)\n",
    "#     df_country['type'].fillna('unknown', inplace=True)\n",
    "#     df_country['type'] = np.where(df_country['type'] == 'residential', df_country['residential_type'], df_country['type'])\n",
    "\n",
    "#     sns.violinplot(data=df_country, x='type', y='error', order=order, hue='Country', inner='quartile', split=True, palette='huel', ax=ax)\n",
    "\n",
    "#     ax.annotate(text='residential\\nbuilding types', xy=(annot_pos[0],-0.26), xycoords='axes fraction', horizontalalignment='center', fontsize=20)\n",
    "#     ax.annotate(text='non-residential\\nbuilding types', xy=(annot_pos[1],-0.26), xycoords='axes fraction', horizontalalignment='center', fontsize=20)\n",
    "#     ax.axvline(3.5, color='grey', linestyle='--', lw=0.75)\n",
    "#     ax.axvline(len(order) - 1.5, color='grey', linestyle='--', lw=0.75)\n",
    "\n",
    "#     ax.set_ylim([-120, 120])\n",
    "#     ax.set_ylabel('Prediction error [years]')\n",
    "#     ax.set_xlabel(None)\n",
    "#     ax.set_xticklabels(['single-family house', 'terraced house', 'multi-family house', 'apartment block', 'commercial', 'agricultural', 'industrial', 'others', 'unknown'])\n",
    "#     plt.xticks(rotation=30)\n",
    "#     plt.title(f'Prediction error for fine-granular building types in {country}')\n",
    "    # plt.savefig(f'../figures/error-by-type-{country}-violin.svg')\n",
    "\n",
    "    # df_country['type'].value_counts()\n",
    "\n",
    "\n",
    "# with visualizations.SubplotManager(figsize=(19, 9)) as ax: # 16,9 for Spain\n",
    "#     order = ['SFH', 'TH', 'MFH', 'AB', 'commercial', 'agricultural', 'industrial', 'others']\n",
    "#     labels = ['single-family house', 'terraced house', 'multi-family house', 'apartment block', 'commercial', 'agricultural', 'industrial', 'others']\n",
    "#     colors = ['thistle', 'lightseagreen']\n",
    "\n",
    "#     df_country = df_test[df_test['Country'].isin(['Spain', 'France'])]\n",
    "\n",
    "#     df_country = harmonize_group_source_types(df_country)\n",
    "#     df_country = df_country[df_country['type'] != 'unknown']\n",
    "#     df_country['type'] = np.where(df_country['type'] == 'residential', df_country['residential_type'], df_country['type'])\n",
    "\n",
    "#     sns.violinplot(data=df_country, x='type', y='error', order=order, hue='Country', inner='quartile', split=True, palette=colors, ax=ax)\n",
    "\n",
    "#     ax.annotate(text='residential\\nbuilding types', xy=(0.25,-0.35), xycoords='axes fraction', horizontalalignment='center', fontsize=20)\n",
    "#     ax.annotate(text='non-residential\\nbuilding types', xy=(0.70,-0.35), xycoords='axes fraction', horizontalalignment='center', fontsize=20)\n",
    "#     ax.axvline(3.5, color='grey', linestyle='--', lw=0.75)\n",
    "\n",
    "#     ax.set_ylim([-120, 120])\n",
    "#     ax.set_ylabel('Prediction error [years]')\n",
    "#     ax.set_xlabel(None)\n",
    "#     shares = df_country.groupby('Country')['type'].value_counts(normalize=True).mul(100).round(1).sort_index(level=0).groupby('type').apply(list).to_dict()\n",
    "#     shares['others'].append(0)\n",
    "#     labels = [f'{labels[i]}\\n({shares[subtype][0]}% | {shares[subtype][1]}%)' for i, subtype in enumerate(order)]\n",
    "#     ax.set_xticklabels(labels)\n",
    "#     # ax.set_aspect(aspect=0.021)\n",
    "#     plt.xticks(rotation=30)\n",
    "#     plt.title(f'Prediction error for fine-granular building types in France and Spain', pad=20)\n",
    "#     plt.savefig(f'../figures/error-by-finegranular-type-violin.svg')\n",
    "\n",
    "#     df_country['type'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "# error_additional_data = pd.read_csv('../results/results-additional-data-23785993-20220622-02-07-11.csv')\n",
    "# error_additional_data_random = pd.read_csv('../results/results-additional-data-nl-regression-inner-city.csv') # old, with only 1 seed\n",
    "\n",
    "paths = {\n",
    "    'Netherlands': {\n",
    "        'Neighborhood cv': '../results/results-additional-data-neighborhood-cv-23839686-20220708-00-59-18.csv',\n",
    "        'Random cv' : '../results/results-additional-data-inner-cities-23855432-20220712-12-06-14.csv',\n",
    "        'Across cities': '../results/results-additional-data-across-cities-23834715-20220707-11-23-51.csv',\n",
    "        'Across countries': '../results/results-additional-data-across-countries-23834715-20220707-11-23-51.csv',\n",
    "    },\n",
    "    'France': {\n",
    "        'Neighborhood cv': '../results/results-additional-data-neighborhood-cv-FRA-23879948-20220720-02-07-55.csv',\n",
    "        'Random cv' : '../results/results-additional-data-inner-cities-FRA-23879948-20220720-02-07-55.csv',\n",
    "        'Across cities': '../results/results-additional-data-across-cities-FRA-23879948-20220720-02-07-55.csv',\n",
    "        'Across countries': '../results/results-additional-data-across-countries-FRA-23879948-20220720-02-07-55.csv',\n",
    "    },\n",
    "}\n",
    "\n",
    "# NLD\n",
    "# error_additional_data_random = pd.read_csv('../results/results-additional-data-inner-cities-23855432-20220712-12-06-14.csv')\n",
    "# error_additional_data_neighborhood = pd.read_csv('../results/results-additional-data-neighborhood-cv-23839686-20220708-00-59-18.csv')\n",
    "# error_additional_data_city = pd.read_csv('../results/results-additional-data-across-cities-23834715-20220707-11-23-51.csv')\n",
    "# error_additional_data_country = pd.read_csv('../results/results-additional-data-across-countries-23834715-20220707-11-23-51.csv')\n",
    "# ylim = (0, 0.9)\n",
    "# bld_per_city = '13,000'\n",
    "# n_cities = 424\n",
    "\n",
    "# # FRA\n",
    "error_additional_data_random = pd.read_csv('../results/results-additional-data-inner-cities-FRA-23879948-20220720-02-07-55.csv')\n",
    "error_additional_data_neighborhood = pd.read_csv('../results/results-additional-data-neighborhood-cv-FRA-23879948-20220720-02-07-55.csv')\n",
    "error_additional_data_city = pd.read_csv('../results/results-additional-data-across-cities-FRA-23879948-20220720-02-07-55.csv')\n",
    "error_additional_data_country = pd.read_csv('../results/results-additional-data-across-countries-FRA-23879948-20220720-02-07-55.csv')\n",
    "ylim = (-0.5, 0.8)\n",
    "bld_per_city = '4,000'\n",
    "n_cities = 2048\n",
    "\n",
    "# Plot individual experiments\n",
    "# visualizations.plot_impact_of_additional_data_on_error(error_additional_data_neighborhood)\n",
    "\n",
    "# Plot R2 of all experiments in single graph\n",
    "linestyles = ['-', '--', ':', '-.']\n",
    "\n",
    "with visualizations.SubplotManager() as ax1:\n",
    "    for i, (name, df) in enumerate({'Local inference (random cv)' : error_additional_data_random, 'Baseline (neighborhood cv)': error_additional_data_neighborhood, 'Generalization across cities': error_additional_data_city, 'Generalization across countries': error_additional_data_country}.items()):\n",
    "\n",
    "        df = df[df['name'] != 'baseline']\n",
    "        df['n_cities'] = df['name'].str.split('-').str[0].astype(int)\n",
    "        df.set_index('n_cities', inplace=True)\n",
    "        df.sort_index(inplace=True)\n",
    "        df.rename(columns={'R2': name}, inplace=True)\n",
    "        df[[name]].plot(ax=ax1, xticks=df.index, xlabel=f'Number of cities used for training (~{bld_per_city} buildings per city on average)', ylabel='$R^2$', ylim=ylim, logx=True, style=linestyles[i], legend=True)\n",
    "        ticks = [2 ** x for x in range(13)][::2]\n",
    "        ax1.set_xticks(ticks, labels=ticks)\n",
    "        ax1.set_yticks(np.arange(ylim[0], ylim[1], 0.2))\n",
    "        ax1.legend(loc=1)\n",
    "\n",
    "        ax1.fill_between(\n",
    "            df.index,\n",
    "            df[name] - df['R2_std'],\n",
    "            df[name] + df['R2_std'],\n",
    "            alpha=0.1,\n",
    "        )\n",
    "\n",
    "    ax1.axvline(n_cities, color='grey', linestyle='--', lw=0.75)\n",
    "    ax1.xaxis.labelpad = 20\n",
    "    ax1.set_title('Impact of additional training data on prediction performance in France', pad=20)\n",
    "    plt.savefig('../figures/additional-data-overview-FRA-inc-other-country.svg')\n",
    "\n",
    "\n",
    "\n",
    "# # Save results as latex table\n",
    "# dfs = []\n",
    "# for country, exp_paths in paths.items():\n",
    "#     for name, path in exp_paths.items():\n",
    "#         df = pd.read_csv(path)\n",
    "#         df['Use case'] = name\n",
    "#         df['Country'] = country\n",
    "#         df['n cities'] = df['name'].str.split('-').str[0].astype(int)\n",
    "#         df.sort_values('n cities', inplace=True)\n",
    "#         dfs.append(df[['Use case', 'Country','n cities', 'R2', 'R2_std', 'MAE', 'MAE_std', 'RMSE', 'RMSE_std']])\n",
    "\n",
    "# error_additional_data = pd.concat(dfs)\n",
    "\n",
    "# pivot_table = error_additional_data.pivot_table(values=['R2', 'MAE', 'RMSE'], index=['Country', 'n cities'], columns=['Use case'], margins=False)\n",
    "# pivot_table = pivot_table.swaplevel(0, 1, axis=1)\n",
    "# # pivot_table.columns = pivot_table.columns.rename('Error', level=1)\n",
    "# pivot_table.columns = pivot_table.columns.rename(None, level=0)\n",
    "# #.reorder_levels(order = [1,0,2], axis=1)\n",
    "# # pivot_table.columns = pd.MultiIndex.from_tuples([(c[1], c[0], c[2]) for c in pivot_table.columns])\n",
    "# pivot_table = pivot_table.sort_index(axis=1, level=0)\n",
    "# new_cols = pivot_table.columns.reindex(list(paths['France'].keys()), level=0)\n",
    "# pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "# new_cols = pivot_table.columns.reindex(['R2', 'MAE', 'RMSE'], level=1)\n",
    "# pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "\n",
    "# def f2(x):\n",
    "#     return '%.2f' % x\n",
    "# def f4(x):\n",
    "#     return '%.4f' % x\n",
    "\n",
    "# # pivot_table.to_latex('../figures/additional-data.tex', index=True, float_format='%.2f', formatters=[f4, f2, f2,f4, f2, f2,f4, f2, f2,f4, f2, f2], caption='Impact of additional data on prediction performance', label='table:additional-data')\n",
    "# pivot_table.to_latex('../figures/additional-data.tex', index=True, float_format='%.2f', caption='Impact of additional data on prediction performance', label='table:additional-data')\n",
    "# pivot_table"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure: Spatial distance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "['name'] did not aggregate successfully. If any error is raised this will raise in a future version of pandas. Drop these columns/ops to avoid this warning.\n",
      "['name'] did not aggregate successfully. If any error is raised this will raise in a future version of pandas. Drop these columns/ops to avoid this warning.\n",
      "['name'] did not aggregate successfully. If any error is raised this will raise in a future version of pandas. Drop these columns/ops to avoid this warning.\n"
     ]
    }
   ],
   "source": [
    "# sp_comparison = pd.read_csv('../results/results-spatial-distance-23839686-20220709-04-14-17.csv')\n",
    "sp_comparison_nl = pd.read_csv('../results/results-spatial-distance-NLD-23855379-20220709-16-42-46.csv')\n",
    "sp_comparison_es = pd.read_csv('../results/results-spatial-distance-ESP-23855379-20220709-16-42-46.csv')\n",
    "sp_comparison_fr = pd.read_csv('../results/results-spatial-distance-FRA-23855432-20220711-10-23-04.csv')\n",
    "visualizations.plot_impact_of_spatial_distance_on_error([sp_comparison_nl, sp_comparison_fr, sp_comparison_es], ['Netherlands', 'France', 'Spain'])\n",
    "plt.savefig('../figures/spatial-distance.svg')\n",
    "\n",
    "# sp_comparison['distance'] = sp_comparison['name'].str.split('-').str[0].astype(int)\n",
    "# # sp_comparison[sp_comparison['R2'].abs() > 1]\n",
    "# # sp_comparison[sp_comparison['name'].str.contains('Flevoland')]\n",
    "# sp_comparison[sp_comparison['name'].str.contains('Gelderland')].sort_values(by='R2')\n",
    "# # sp_comparison[sp_comparison['distance'] == 25]\n",
    "# sp_comparison = sp_comparison.groupby('distance').agg(['mean', 'std'])\n",
    "# sp_comparison.sort_values(by=('R2', 'mean'))\n",
    "# sp_comparison[[('RMSE', 'mean'), ('MAE', 'mean')]]\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 771,
   "metadata": {},
   "outputs": [],
   "source": [
    "# util from https://stackoverflow.com/questions/31908982/python-matplotlib-multi-color-legend-entry\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.collections import PatchCollection\n",
    "\n",
    "# define an object that will be used by the legend\n",
    "class MulticolorPatch(object):\n",
    "    def __init__(self, colors):\n",
    "        self.colors = colors\n",
    "        \n",
    "# define a handler for the MulticolorPatch object\n",
    "class MulticolorPatchHandler(object):\n",
    "    def legend_artist(self, legend, orig_handle, fontsize, handlebox):\n",
    "        width, height = handlebox.width, handlebox.height\n",
    "        patches = []\n",
    "        for i, c in enumerate(orig_handle.colors):\n",
    "            patches.append(plt.Rectangle([width/len(orig_handle.colors) * i - handlebox.xdescent, \n",
    "                                          -handlebox.ydescent],\n",
    "                           width / len(orig_handle.colors),\n",
    "                           height, \n",
    "                           facecolor=c, \n",
    "                           edgecolor='none'))\n",
    "\n",
    "        patch = PatchCollection(patches,match_original=True)\n",
    "\n",
    "        handlebox.add_artist(patch)\n",
    "        return patch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 929,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "You are modifying the state of a globally registered colormap. This has been deprecated since 3.3 and in 3.6, you will not be able to modify a registered colormap in-place. To remove this warning, you can make a copy of the colormap first. cmap = mpl.cm.get_cmap(\"Pastel2_r\").copy()\n"
     ]
    }
   ],
   "source": [
    "with open(os.path.join('..', 'metadata', f'experiment-spatial-distance-cities-FRA.json'), 'r', encoding='utf-8') as f:\n",
    "    cities = json.load(f)\n",
    "\n",
    "distance_groups = {city: dis for dis, (train_cities, _) in enumerate(cities['Allier']) for city in train_cities}\n",
    "departments = geometry.to_gdf(gadm_france)\n",
    "departments['dis'] = departments['NAME_4'].map(distance_groups).astype(float) * 50 + 50\n",
    "test_region = departments[departments['NAME_2'] == 'Allier'].dissolve('GID_2')\n",
    "\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax1:\n",
    "    c_test = 'violet'\n",
    "    c_train = 'purple'\n",
    "    c_rings = 'violet'\n",
    "    cmap = matplotlib.colors.LinearSegmentedColormap.from_list(\"\", [c_train, 'lightblue'])\n",
    "\n",
    "    departments.boundary.plot(ax=ax1, facecolor=None, edgecolor='grey', alpha=0.4, linewidth=0.05)\n",
    "    # departments.dissolve('GID_2').boundary.simplify(1000).plot(ax=ax1, color='grey', alpha=0.1, linewidth=0.5)\n",
    "    departments.dissolve('GID_0').boundary.simplify(1000).plot(ax=ax1, facecolor='none', edgecolor='grey', alpha=0.5, linewidth=0.6)\n",
    "\n",
    "    cmap = matplotlib.cm.get_cmap('Pastel2_r')\n",
    "    colors = list(reversed(cmap.colors + ['pink']))\n",
    "    # reversed to ensure colors are not overlapping and ending up all as different gray tones\n",
    "    for idx, buffer_size in enumerate(reversed(range(100000, 550000, 50000))):\n",
    "        buffer = test_region['geometry'].centroid.buffer(buffer_size)\n",
    "        departments[departments.intersects(buffer.iloc[0])].plot(ax=ax1, facecolor=colors[idx], edgecolor='none', linewidth=0.65, alpha=0.2)\n",
    "        buffer.boundary.plot(ax=ax1, facecolor='none', edgecolor=c_rings, linestyle='dashed', linewidth=1.5, alpha=0.25+idx/12)\n",
    "        # departments[departments.intersects(buffer.iloc[0])].apply(lambda x: ax1.annotate(text=x[f'NAME_4'], xy=x.geometry.centroid.coords[0], ha='center'), axis=1)\n",
    "\n",
    "    test_region.plot(ax=ax1, color=c_test, edgecolor=None, linewidth=0.5, alpha=0.9)\n",
    "    cmap.set_over('pink')\n",
    "    departments[departments['NAME_4'] != 'Saint-Savin'].dropna(subset='dis').plot(ax=ax1, column='dis', alpha=1, cmap=cmap, vmax=400, linewidth=1, legend=True, legend_kwds={'shrink': 0.63, 'orientation': 'vertical', 'pad': 0,  'label': 'Distance to test region [km]'})\n",
    "\n",
    "\n",
    "    plt.axis('off')\n",
    "    from matplotlib.lines import Line2D\n",
    "    import matplotlib.patches as mpatches\n",
    "\n",
    "    handles = [MulticolorPatch(colors), MulticolorPatch([c_test]), Line2D([0], [0], color=c_rings, linestyle='dashed', linewidth=1.5)]\n",
    "    labels = ['Train regions', 'Test region', 'Distance rings']\n",
    "    ax1.legend(loc='upper right', fancybox=True, handles=handles, labels=labels, fontsize=12, handler_map={MulticolorPatch: MulticolorPatchHandler()})\n",
    "\n",
    "    plt.savefig('../figures/spatial-distance-visualization.png')\n",
    "\n",
    "# Visualization of determining 9 regional groups for the departmant of Allier in France.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 856,
   "metadata": {},
   "outputs": [],
   "source": [
    "# mpatches.Patch(facecolor=)\n",
    "for color in matplotlib.cm.get_cmap('Pastel2').colors:\n",
    "    with visualizations.SubplotManager() as ax:\n",
    "        ax.plot([1,2,3,4], color=color)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 757,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((0.7019607843137254, 0.8862745098039215, 0.803921568627451),\n",
       " (0.9921568627450981, 0.803921568627451, 0.6745098039215687),\n",
       " (0.796078431372549, 0.8352941176470589, 0.9098039215686274),\n",
       " (0.9568627450980393, 0.792156862745098, 0.8941176470588236),\n",
       " (0.9019607843137255, 0.9607843137254902, 0.788235294117647),\n",
       " (1.0, 0.9490196078431372, 0.6823529411764706),\n",
       " (0.9450980392156862, 0.8862745098039215, 0.8),\n",
       " (0.8, 0.8, 0.8))"
      ]
     },
     "execution_count": 757,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "handles = [mpatches.Patch(color) for color in matplotlib.cm.get_cmap('Pastel2').colors]\n",
    "handles += mpatches.Patch(color=c_test, label='Test region')\n",
    "labels =[''] * len(handles)\n",
    "labels[-2] = 'Train regions'\n",
    "labels[-1] = 'Test region'\n",
    "ax1.legend(loc='upper right', fancybox=True, handles=handles, labels=labels, fontsize=12)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 685,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>NAME_0</th>\n",
       "      <th>NAME_1</th>\n",
       "      <th>NAME_2</th>\n",
       "      <th>NAME_3</th>\n",
       "      <th>NAME_4</th>\n",
       "      <th>GID_0</th>\n",
       "      <th>GID_1</th>\n",
       "      <th>GID_2</th>\n",
       "      <th>GID_3</th>\n",
       "      <th>GID_4</th>\n",
       "      <th>VARNAME_4</th>\n",
       "      <th>TYPE_4</th>\n",
       "      <th>ENGTYPE_4</th>\n",
       "      <th>CC_4</th>\n",
       "      <th>geometry</th>\n",
       "      <th>DEGURBA</th>\n",
       "      <th>dis</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Ain</td>\n",
       "      <td>Belley</td>\n",
       "      <td>Ambérieu-en-Bugey</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.1_1</td>\n",
       "      <td>FRA.1.1.1_1</td>\n",
       "      <td>FRA.1.1.1.1_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3952485.219 2547275.335, 3952754.225...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Ain</td>\n",
       "      <td>Belley</td>\n",
       "      <td>Belley</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.1_1</td>\n",
       "      <td>FRA.1.1.1_1</td>\n",
       "      <td>FRA.1.1.1.2_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3984571.942 2513227.555, 3983634.519...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Ain</td>\n",
       "      <td>Belley</td>\n",
       "      <td>Champagne-en-Valromey</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.1_1</td>\n",
       "      <td>FRA.1.1.1_1</td>\n",
       "      <td>FRA.1.1.1.3_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3991409.530 2539244.014, 3992261.974...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Ain</td>\n",
       "      <td>Belley</td>\n",
       "      <td>Hauteville-Lompnes</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.1_1</td>\n",
       "      <td>FRA.1.1.1_1</td>\n",
       "      <td>FRA.1.1.1.4_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3978689.098 2537916.471, 3975920.723...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Ain</td>\n",
       "      <td>Belley</td>\n",
       "      <td>Lagnieu</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.1_1</td>\n",
       "      <td>FRA.1.1.1_1</td>\n",
       "      <td>FRA.1.1.1.5_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3966127.201 2534993.813, 3965990.279...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3300</th>\n",
       "      <td>France</td>\n",
       "      <td>Île-de-France</td>\n",
       "      <td>Yvelines</td>\n",
       "      <td>Versailles</td>\n",
       "      <td>Saint-Cyr-l'École</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.8_1</td>\n",
       "      <td>FRA.8.8_1</td>\n",
       "      <td>FRA.8.8.4_1</td>\n",
       "      <td>FRA.8.8.4.4_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3738058.012 2885390.475, 3736575.155...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3301</th>\n",
       "      <td>France</td>\n",
       "      <td>Île-de-France</td>\n",
       "      <td>Yvelines</td>\n",
       "      <td>Versailles</td>\n",
       "      <td>Trappes</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.8_1</td>\n",
       "      <td>FRA.8.8_1</td>\n",
       "      <td>FRA.8.8.4_1</td>\n",
       "      <td>FRA.8.8.4.5_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3733387.020 2886037.449, 3733427.211...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3302</th>\n",
       "      <td>France</td>\n",
       "      <td>Île-de-France</td>\n",
       "      <td>Yvelines</td>\n",
       "      <td>Versailles</td>\n",
       "      <td>Versailles</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.8_1</td>\n",
       "      <td>FRA.8.8_1</td>\n",
       "      <td>FRA.8.8.4_1</td>\n",
       "      <td>FRA.8.8.4.8_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3741150.525 2875898.862, 3741021.743...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3303</th>\n",
       "      <td>France</td>\n",
       "      <td>Île-de-France</td>\n",
       "      <td>Yvelines</td>\n",
       "      <td>Versailles</td>\n",
       "      <td>Viroflay</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.8_1</td>\n",
       "      <td>FRA.8.8_1</td>\n",
       "      <td>FRA.8.8.4_1</td>\n",
       "      <td>FRA.8.8.4.9_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3746554.126 2883296.084, 3746112.981...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3304</th>\n",
       "      <td>France</td>\n",
       "      <td>Île-de-France</td>\n",
       "      <td>Yvelines</td>\n",
       "      <td>Versailles</td>\n",
       "      <td>Vélizy-Villacoublay</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.8_1</td>\n",
       "      <td>FRA.8.8_1</td>\n",
       "      <td>FRA.8.8.4_1</td>\n",
       "      <td>FRA.8.8.4.6_1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Canton</td>\n",
       "      <td>Cantons</td>\n",
       "      <td>NaN</td>\n",
       "      <td>POLYGON ((3750683.679 2881539.897, 3750691.639...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3305 rows × 17 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      NAME_0                NAME_1    NAME_2      NAME_3  \\\n",
       "0     France  Auvergne-Rhône-Alpes       Ain      Belley   \n",
       "1     France  Auvergne-Rhône-Alpes       Ain      Belley   \n",
       "2     France  Auvergne-Rhône-Alpes       Ain      Belley   \n",
       "3     France  Auvergne-Rhône-Alpes       Ain      Belley   \n",
       "4     France  Auvergne-Rhône-Alpes       Ain      Belley   \n",
       "...      ...                   ...       ...         ...   \n",
       "3300  France         Île-de-France  Yvelines  Versailles   \n",
       "3301  France         Île-de-France  Yvelines  Versailles   \n",
       "3302  France         Île-de-France  Yvelines  Versailles   \n",
       "3303  France         Île-de-France  Yvelines  Versailles   \n",
       "3304  France         Île-de-France  Yvelines  Versailles   \n",
       "\n",
       "                     NAME_4 GID_0    GID_1      GID_2        GID_3  \\\n",
       "0         Ambérieu-en-Bugey   FRA  FRA.1_1  FRA.1.1_1  FRA.1.1.1_1   \n",
       "1                    Belley   FRA  FRA.1_1  FRA.1.1_1  FRA.1.1.1_1   \n",
       "2     Champagne-en-Valromey   FRA  FRA.1_1  FRA.1.1_1  FRA.1.1.1_1   \n",
       "3        Hauteville-Lompnes   FRA  FRA.1_1  FRA.1.1_1  FRA.1.1.1_1   \n",
       "4                   Lagnieu   FRA  FRA.1_1  FRA.1.1_1  FRA.1.1.1_1   \n",
       "...                     ...   ...      ...        ...          ...   \n",
       "3300      Saint-Cyr-l'École   FRA  FRA.8_1  FRA.8.8_1  FRA.8.8.4_1   \n",
       "3301                Trappes   FRA  FRA.8_1  FRA.8.8_1  FRA.8.8.4_1   \n",
       "3302             Versailles   FRA  FRA.8_1  FRA.8.8_1  FRA.8.8.4_1   \n",
       "3303               Viroflay   FRA  FRA.8_1  FRA.8.8_1  FRA.8.8.4_1   \n",
       "3304    Vélizy-Villacoublay   FRA  FRA.8_1  FRA.8.8_1  FRA.8.8.4_1   \n",
       "\n",
       "              GID_4  VARNAME_4  TYPE_4 ENGTYPE_4  CC_4  \\\n",
       "0     FRA.1.1.1.1_1        NaN  Canton   Cantons   NaN   \n",
       "1     FRA.1.1.1.2_1        NaN  Canton   Cantons   NaN   \n",
       "2     FRA.1.1.1.3_1        NaN  Canton   Cantons   NaN   \n",
       "3     FRA.1.1.1.4_1        NaN  Canton   Cantons   NaN   \n",
       "4     FRA.1.1.1.5_1        NaN  Canton   Cantons   NaN   \n",
       "...             ...        ...     ...       ...   ...   \n",
       "3300  FRA.8.8.4.4_1        NaN  Canton   Cantons   NaN   \n",
       "3301  FRA.8.8.4.5_1        NaN  Canton   Cantons   NaN   \n",
       "3302  FRA.8.8.4.8_1        NaN  Canton   Cantons   NaN   \n",
       "3303  FRA.8.8.4.9_1        NaN  Canton   Cantons   NaN   \n",
       "3304  FRA.8.8.4.6_1        NaN  Canton   Cantons   NaN   \n",
       "\n",
       "                                               geometry  DEGURBA  dis  \n",
       "0     POLYGON ((3952485.219 2547275.335, 3952754.225...      3.0  NaN  \n",
       "1     POLYGON ((3984571.942 2513227.555, 3983634.519...      3.0  NaN  \n",
       "2     POLYGON ((3991409.530 2539244.014, 3992261.974...      3.0  NaN  \n",
       "3     POLYGON ((3978689.098 2537916.471, 3975920.723...      3.0  NaN  \n",
       "4     POLYGON ((3966127.201 2534993.813, 3965990.279...      3.0  NaN  \n",
       "...                                                 ...      ...  ...  \n",
       "3300  POLYGON ((3738058.012 2885390.475, 3736575.155...      2.0  NaN  \n",
       "3301  POLYGON ((3733387.020 2886037.449, 3733427.211...      1.0  NaN  \n",
       "3302  POLYGON ((3741150.525 2875898.862, 3741021.743...      1.0  NaN  \n",
       "3303  POLYGON ((3746554.126 2883296.084, 3746112.981...      1.0  NaN  \n",
       "3304  POLYGON ((3750683.679 2881539.897, 3750691.639...      2.0  NaN  \n",
       "\n",
       "[3305 rows x 17 columns]"
      ]
     },
     "execution_count": 685,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "departments"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prediction error figures"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "regressor_fts_importance_fr = AgePredictor.load('../results/model-FRA-neighborhood-cv-fts-importance.pkl')\n",
    "regressor_fts_importance_nl = AgePredictor.load('../results/model-NLD-neighborhood-cv-fts-importance.pkl')\n",
    "regressor_fts_importance_es = AgePredictor.load('../results/model-ESP-neighborhood-cv-fts-importance.pkl')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# regressor_nl_random_cv = AgePredictor.load('../results/model-NLD-random-cv.pkl')\n",
    "# regressor_nl_block_cv = AgePredictor.load('../results/model-NLD-block-cv.pkl')\n",
    "# regressor_nl_neighhorhood_cv = AgePredictor.load('../results/model-NLD-neighborhood-cv-fts-importance.pkl')\n",
    "# regressor_nl_city_cv = AgePredictor.load('../results/model-NLD-city-cv.pkl')\n",
    "regressor_country_cv = AgePredictor.load('../results/model-country-cv.pkl')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Figure: Prediction error over construction periods"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1917805\n",
      "9596614\n"
     ]
    }
   ],
   "source": [
    "print(len(regressor_fts_importance_fr.X_test))\n",
    "print(len(regressor_fts_importance_fr.aux_vars_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAuYAAALECAYAAABE/fVuAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAADdxElEQVR4nOz9d3xc5bXo/3/2dI2kUbMt2ZZlWy5jW24Y3ADTDYcaMIFAHCAhIZXcfHPOyQ25v5x7bhrk5KRzkkACCQkkYIoNmFAMNrji3m15VGz13sto+v79MZaQbNlWmZm9Z7zer5deWDN79rMkJM3az17PehRVVRFCCCGEEEJoy6B1AEIIIYQQQghJzIUQQgghhNAFScyFEEIIIYTQAUnMhRBCCCGE0AFJzIUQQgghhNABk9YB6IHT6bQCi4FaIKhxOEIIIYQQInEZgfHAHpfL5e3/hCTmYYuBrVoHIYQQQgghLhorgG39H5DEPKwW4O9//zs5OTlaxyKEEEIIIRJUXV0dq1evhtP5Z3+SmIcFAXJycsjNzdU6FiGEEEIIkfjOKp+WxZ9CCCGEEELogCTmQgghhBBC6IAk5kIIIYQQQuiAJOZCCCGEEELogCz+FEIIIYS4gI6ODhoaGvD7/VqHInTObDYzbtw4HA7HsF8ribkQQgghxHl0dHRQX1/PxIkTSUpKQlEUrUMSOqWqKj09PVRXVwMMOzmXUhYhhBBCiPNoaGhg4sSJ2O12ScrFeSmKgt1uZ+LEiTQ0NAz79ZKYCyGEEEKch9/vJykpSeswRBxJSkoaUdmTJOZCCCGEEBcgM+ViOEb68yKJuRBCCCGEEDogibkQQgghhBjg4YcfpqWlReswLjqSmAshhBBCiAG2b9+udQgXJUnMhRBCCCFEn+9973sArFixAqfTSSgUAqCnp4fly5fT3NzMddddx+OPP86qVatYuXIl//jHP/pev2nTJu655x7uvPNO7rvvPg4cOKDJ1xGPpI+5EEIIIYTo88QTT7B27Vq2bt3KF77wBbZu3crVV1/NP//5T5YvX05WVhYAHo+H1157jfr6eu68804uvfRSrFYrv/rVr/jb3/5GRkYGxcXFfOELX2DDhg3Y7XaNvzL9k8RcCCGEEEIMavXq1bz88stcffXVrFmzhv/9v/9333Of/exnURSFnJwcVqxYwfbt27FarTQ0NPD5z3++7zhFUaioqGDWrFkafAXxRRJzIYQQQggxqNtvv51f/vKX7Ny5E7fbzeLFi/ueM5k+SSNDoRAGg4FQKMTy5cv59a9/3fdcbW0t48aNi2XYcUtqzIUQQgghxABGo5FAIEBSUhJ33HEH/+f//B/uu+++Ace8/vrrANTU1LB9+3auuuoqli1bxvbt2yktLQVg8+bN3HHHHXi93lh/CXFJZsyFEEIIIcQAK1eu5LOf/Sy///3vWbVqFS+//DJ33nnngGOqqqpYtWoVHo+H73//++Tn5wPwwx/+kH/9139FVVVMJhN/+MMfpL58iCQxF0IIIYQQA/zmN78BQFVV/vSnP/GpT32K1NTUAcd88YtfZN68eWe99uabb+bmm2+OSZyJRhJzIYQQQggxqOuvv57MzEz+8Ic/aB3KRUEScyGEEEIIMahNmzYN63ExOrL4UwghhBBCCB2QxFwIIYQQQggdkMRcCCGEEEIIHZDEXAghhBBCCB2QxFwIIYQQQggdkMRcxL36+nreeOMNrcMQQgghhBgVScxF3HvmmWdYs2YNbW1tWocihBBCiFF64IEH+NWvfjXoc4FAAKfTya5duyIy1rvvvktjYyMATz75JPfff/+QXvfYY4/x7//+7xGJoT/pYy7iXnV1NRDenUwIIYSIhR/99AlaOtpiNl6mI53/eOx7MRtPS08++SRmsznq41RXV/Otb32LDRs2APDwww/zwAMPRH3c85HEXAghhBBimFo62jCsKIjdeFuPxWwsraWnp8dknDMn9JKTk2My7vlIKYsQQgghRIKpqqrC6XTy3nvvsXLlSubNm8cjjzxCS0sLAAcOHOD+++9n4cKFXHfddfz973/ve+1jjz3Gj3/8Y/71X/+VhQsXctVVV7F27doLjvW73/2OxYsX873vhWf2P/jgA2699VYWLFjAXXfdxZYtW/pe43K5WL16NQsXLuSKK67gpz/9KYFAADi7lOV//ud/WL58OcuWLWPdunUDxvb5fPzkJz9h2bJlLF26lG9961s0NTUN6Xtw/fXXA3DjjTeydu3as0pZXnvtNW6++Wbmzp3L0qVL+c///M++GKNFEnMhhBBCiAT19NNP8/Of/5wXXniBY8eO8eyzz1JaWspDDz3E4sWLWbduHd/85jf57//+b955552+17300kvMnj2b9evXc9NNN/H//t//u+Barr179/Laa6/x5S9/mRMnTvCd73yHRx55hPXr13Pvvffy6KOPUlhYCMB3vvMd8vPzWb9+Pb/+9a954403ePXVV88655o1a/jb3/7G448/zl/+8hdee+21Ac//8pe/5ODBgzz99NM8//zzqKrKV77ylQGz4YN9DwBeeeWVvjFuueWWs76WH/zgB3z729/mvffe4wc/+AFr167tK3uJFillEUIIIYRIUI8++igLFiwA4Pbbb+fIkSN9Cyj/9V//FYCpU6dSWlrKM888w8033wzAzJkzeeSRRwD41re+xd/+9jeKi4tZvHjxOcd68MEHycvLA8KJ9913382dd94JQF5eHocPH+b555/n8ccfp7q6mmuuuYYJEyYwadIk/vSnPw1awvLyyy/zwAMPcO211wLwwx/+kNtvvx2Anp4eXnjhBV5++WXmzJkDwM9+9jOWLl3Kvn37yMnJOef3ACAzMxOAjIwMbDbbgHFtNhs/+clPuPHGGwGYOHEif/nLXygpKRnS932kJDEXQgghhEhQvYkyQEpKCoFAgNLS0r5Etdcll1wyoJxl0qRJA14H4Y4oe/fu7UvYAb7yla9w2223AeHktVdpaSlFRUUDZrj9fj/z588H4Gtf+xq/+MUvWLNmDVdddRW33norc+fOPSv+0tJSvvrVr/Z9PnPmTKxWKwCVlZX4/X5Wr1494DVer5dTp071JeaDfQ8uZO7cudhsNn77299SUlKCy+WivLycZcuWXfC1oyGJuRBCCCFEgjqzu4mqqmfNDgOEQiGCweA5X9f72rlz5/L666/3PZaWlkZXVxdAX8IMEAwG+eIXv8iqVasGnMNisQDwpS99iVtuuYWNGzfy0Ucf8fWvf52vfe1rfPOb3xx03P6MRmPfGADPP/88qampA47JzMykvb39nN+DC9m6dStf//rXufPOO1mxYgXf+MY3+MEPfnDB142W1JiLuBcKhbQOQQghhIgb+fn5HDp0aMBjBw4cYOrUqRd8rc1mY/LkyX0f5+qgMnXqVCorKwcc+8Ybb/D+++/j9Xr58Y9/jKIoPPDAAzz77LM8+uijvP3222edZ8aMGX2lJwDl5eW43W4gPKtvNBppbW3tGyMzM5Mnnniir5Xy+SiKcs7nXnnlFe666y5+9KMfcc899zBt2jQqKioueM7RksRcxL3eX9DOzk6NIxFCCCH077Of/SxFRUX88pe/5NSpU7z++uv84x//4HOf+1zExvj85z/Pu+++y3PPPUd5eTkvvvgiTz31FJMnT8ZqtbJ//35+9KMfUVpaisvlYsuWLRQUnN1+cvXq1bzwwgu88847FBUV8f3vfx+DIZy+pqSkcM899/CjH/2Ijz/+mNLSUr773e9SVFTElClTLhij3W4H4MSJE3R3dw94Lj09nQMHDnDixAmKi4t57LHHaGxsxOfzjf6bcx6SmIu413srq/8tOCGEEEIMLicnh6effppt27Zx++238/vf/57HHnuMe+65J2JjLFy4kJ///Oe8/PLL3HrrrTz33HM8/vjjXH311QD86le/wuv1cu+99/LZz36W3Nxc/uM//uOs89x5551861vf4ic/+QmrV6/mmmuuGdBv/LHHHuOKK67g29/+Np/+9Kfxer08++yzg5brnCkjI4NVq1bxb//2b30dWno9+uijjBs3jvvuu48vfOELmM1mVq9ezfHjx0f5nTk/RXZLBKfTOQU4tXHjRnJzc7UORwzT6tWrUVWVn/zkJ0O6DSeEEEIMR2FhIbNnzx7wmOz8KS5ksJ8bCPdXP91DfarL5Srr/5ws/hRCCCGEGCZJkkU0SCmLEEIIIYQQOiCJuRBCCCGEEDogibkQQgghhBA6IIm5EEIIIYQQOiCJuUgY59soQAghhBBC7yQxF0IIIYQQQgckMRdCCCGEEEIHJDEXQgghhBBCByQxFwlDasyFEEKIyHvggQf41a9+NehzgUAAp9PJrl27IjLWu+++S2NjIwBPPvkk999//5Be99hjj/Hv//7vEYlBS7LzpxBCCCHEMP3i8R/R2dYSs/FS0zP5t//zHzEbr78nn3wSs9kc9XGqq6v51re+xYYNGwB4+OGHeeCBB6I+rp5IYi6EEEIIMUydbS18fWr0k9Vevz8Vu4uAM6Wnp8dkHFVVB3yenJwck3H1REpZhBBCCCESTFVVFU6nk9/97ncsXryY733ve3zwwQfceuutLFiwgLvuuostW7b0He9yuVi9ejULFy7kiiuu4Kc//SmBQAA4u5Tlf/7nf1i+fDnLli1j3bp1A8b1+Xz85Cc/YdmyZSxdupRvfetbNDU1DYjpvffeY+XKlcybN49HHnmElpbwRcf1118PwI033sjatWvPKmV57bXXuPnmm5k7dy5Lly7lP//zP/tiTBSSmAshhBBCJKi9e/fy2muv8dBDD/Gd73yHRx55hPXr13Pvvffy6KOPUlhYCMB3vvMd8vPzWb9+Pb/+9a954403ePXVV88635o1a/jb3/7G448/zl/+8hdee+21Ac//8pe/5ODBgzz99NM8//zzqKrKV77ylQGz4U8//TQ///nPeeGFFzh27BjPPvssAK+88krfGLfccstZX8cPfvADvv3tb/Pee+/xgx/8gLVr1/aVvSQKKWURQgghhEhQDz74IHl5eXznO9/h7rvv5s477wQgLy+Pw4cP8/zzz/P4449TXV3NNddcw4QJE5g0aRJ/+tOfBi1hefnll3nggQe49tprAfjhD3/I7bffDkBPTw8vvPACL7/8MnPmzAHgZz/7GUuXLmXfvn3k5OQA8Oijj7JgwQIAbr/9do4cOQJAZmYmABkZGdhstgHj2mw2fvKTn3DjjTcCMHHiRP7yl79QUlISwe+W9iQxF0IIIYRIUBMnTgSgtLSUoqKiATPcfr+f+fPnA/C1r32NX/ziF6xZs4arrrqKW2+9lblz5551vtLSUr761a/2fT5z5kysVisAlZWV+P1+Vq9ePeA1Xq+XU6dO9SXmeXl5fc+lpKQMqRxl7ty52Gw2fvvb31JSUoLL5aK8vJxly5YN9VsRFyQxF0IIIYRIUL1JczAY5Itf/CKrVq0a8LzFYgHgS1/6ErfccgsbN27ko48+4utf/zpf+9rX+OY3v3nWOc9cpGk0GvvGAHj++edJTU0dcExmZibt7e0AZ3V4OfN8g9m6dStf//rXufPOO1mxYgXf+MY3+MEPfnDB18UbqTEXCUP6mAshhBCDmzp1KpWVlUyePLnv44033uD999/H6/Xy4x//GEVReOCBB3j22Wd59NFHefvtt886z4wZM/pKTwDKy8txu90ATJo0CaPRSGtra98YmZmZPPHEE1RXV18wxvO9j7/yyivcdddd/OhHP+Kee+5h2rRpVFRUjOA7oW+SmIuEIYm5EEIIMbjPf/7zvPvuuzz33HOUl5fz4osv8tRTTzF58mSsViv79+/nRz/6EaWlpbhcLrZs2UJBQcFZ51m9ejUvvPAC77zzDkVFRXz/+9/HYAinkykpKdxzzz386Ec/4uOPP6a0tJTvfve7FBUVMWXKlAvGaLfbAThx4gTd3d0DnktPT+fAgQOcOHGC4uJiHnvsMRobG/H5fKP/5uiIJOZCCCGEEAlu4cKF/PznP+fll1/m1ltv5bnnnuPxxx/n6quvBuBXv/oVXq+Xe++9l89+9rPk5ubyH/9x9oZGd955J9/61rf4yU9+wurVq7nmmmsG9Bt/7LHHuOKKK/j2t7/Npz/9abxeL88+++xZizkHk5GRwapVq/i3f/u3vg4tvR599FHGjRvHfffdxxe+8AXMZjOrV6/m+PHjo/zO6IsylLqeROd0OqcApzZu3Ehubq7W4YhhWr16Naqq8tOf/nTAghIhhBAiEgoLC5k9e/aAxy6mnT/FyAz2cwPhfu6ne7ZPdblcZf2fk8WfImFIKYsQQohYkSRZRIOUsgghhBBCCKEDkpiLuCflWEIIIYRIBJKYCyGEEEIIoQOSmAshhBBCXIDcnRXDMdKfF0nMhRBCCCHOw2w209PTo3UYIo709PSctcPpUEhiLhKGzGYIIYSIhnHjxlFdXY3b7Zb3GnFeqqridruprq5m3Lhxw369tEsUQgghhDgPh8MBQE1NDX6/X+NohN6ZzWays7P7fm6GQ5eJudPptAL7gP/P5XJ9cPqxTOBp4CagBfhPl8v1136vWQA8BSwACoGvulyuPbGOXWhHZjGEEEJEi8PhGFGiJcRw6K6Uxel02oAXgYIznnoOyAKuAH4IPO10Oi8//Zpk4B1gJ3ApsBX4p9PpTI1R2EIHJDEXQgghRDzTVWLudDrnEE6up53x+DTgduDLLpfriMvl+jPwAvD104d8BvAD/+ZyuQqBbwPtpx8XFwlJzIUQQggRz3SVmANXAx8Cy894fClQ63K5Svo9tq3fccuA7S6XKwTgcrlUYPsg5xEJLBQKaR2CEEIIIcSI6arG3OVy/aH3306ns/9T44GaMw6vB3L7Pe8a5PmFkY1Q6JnMmAshhBAinultxvxc7ID3jMe8gMXpdCrned4ag9iETsiMuRBCCCHiWbwk5h7OTrKtQM/pspVzPe+OQWxCJyQxF0IIIUQ8i5fEvBrIOeOxHKB2iM+Li4CUsgghhBAinsVLYr4TmOh0Oqf0e+zK04/3Pn/56bIWTv/3in7Pi4uAzJgLIYQQIp7FRWLucrlOAu8Bf3M6nfOdTucXgNXA/5w+5FUgBXjydMvFXwKpwEtaxCu0EQwGtQ5BCCGEEGLE4iIxP+1BoA3YBfxf4Esul+tjAJfL1QHcClwO7Cc8W36Ly+Xq1CZUoQWZMRdCCCFEPNNVu8T+XC6XcsbnDcAd5zl+D7Ao2nEJ/ZIZcyGEEELEs3iaMRfivCQxF0IIIUQ8k8RcJAwpZRFCCCFEPJPEXCQMmTEXQgghRDyTxFwkjEAgoHUIQgghhBAjJom5SBiSmAshhBAinkliLhKG3+/XOgQhhBBCiBGTxFwkDJkxF0IIIUQ8k8RcJAyZMRdCCCFEPJPEXCQMr9erdQhCCCGEECMmiblIGD6fT+sQhBBCCCFGTBJzEdf6byokibkQQggh4pkk5iKu9U/GpZRFCCGEEPFMEnMR1/on5h6PR8NIhBBCCCFGRxJzEdf6z5LLjLkQQggh4pkk5iKuyYy5EEIIIRKFJOYirvWfJe/p6dEwEiGEEEKI0ZHEXMS1/om5RxJzIYQQQsQxScxFXOtfvtLjkcRcCCGEEPFLEnMR1wbMmHtk8acQQggh4pck5iKu9Z8xD4VCAzYcEkIIIYSIJ5KYi7jmdrsHfC4LQIUQQggRryQxF3HtzMS8u7tbo0iEEEIIIUZHEnMR185MxCUxF0IIIUS8ksRcxLUzE/Guri6NIhFCCCGEGB1JzEVc6+zsPO/nQgghhBDxQhJzEdc6JDEXQgghRIKQxFzENZkxF0IIIUSikMRcxLWOjva+fxsNCu3t7ec5WgghhBBCvyQxF3Grp6cHn9eHYjIC4LAYJTEXQgghRNySxFzErdbWVgAUswkAh9nQ95gQQgghRLyRxFzErba2NgAUc++MuYG21hYNIxJCCCGEGDlJzEXcamkJJ+GK2Qz0lrJ0oKqqlmEJIYQQQoyIJOYibjU1NQFgsIQT8zSLkUAwKJsMCSGEECIuSWIu4lZjYyOmJBuKUQHCM+aA1JkLIYQQIi5JYi7iVmNjI4Zka9/naZKYCyGEECKOSWIu4lZ9QwOG5KS+z9Os4cS8t/ZcCCGEECKeSGIu4lIoFKKluRljSr/E3GJAAZqbm7ULTAghhBBihCQxF3Gpra2NYDA4IDE3KgqpVpPMmAshhBAiLkliLuJSY2MjAMZ+pSwA6RaDzJgLIYQQIi5JYi7iUm9ibkoZmJhnWIw0n35OCCGEECKeSGIu4lJDQwPAgFIWgAybkabmZtlkSAghhBBxRxJzEZcaGhow2W0oRuOAxzOsRnx+Px0dHRpFJoQQQggxMpKYi7jU0DiwVWKvTJsJ+GRXUCGEEEKIeCGJuYhLdXX1Z5WxQHjGHKRlohBCCCHijyTmIu54PB7aWlsxOpLPei7DKjPmQgghhIhPkpiLuFNXVweAKe3sxNxuUrAYpWWiEEIIIeKPJOYi7vQl5qn2s55TFIUMm0lmzIUQQggRdyQxF3GnqqoKFAXTIKUsABkWhSbpZS6EEEKIOCOJuYg7lZWVmFPtKCbjoM9nWk19fc6FEEIIIeKFJOYi7pSVl2NMTznn85k2I91uN263O4ZRCSGEEEKMjiTmIq54vV4aGxsxZaSe8xjpZS6EEEKIeCSJuYgrlZWVoKqY08+dmI89nZj3LhIVQgghhIgHkpiLuFJWVgaAKdNxzmPGJkliLoQQQoj4I4m5iCtlZWUYrRaMybZzHmMzGXBYTdTW1sYwMiGEEEKI0ZHEXMSVU2WnMGakoCjKeY8bazNSU1MTo6iEEEIIIUZPEnMRN4LBIJWVVZgzzl3G0mtckolaScyFEEIIEUckMRdxo7a2loDfjynz3As/e2XbTXR1d9PR0RGDyIQQQgghRk8ScxE3KioqAIY8Yw5InbkQQggh4oYk5iJulJeXoxgMmBzJFzw2224GkDrzMzQ1NREMBrUOQwghhBCDkMRcxI2y8nJMaSkoxgv/2GZYjZiNBqqrq2MQWXzo6enhf/2v/8XatWu1DkUIIYQQg5DEXMSN8vIyTBkpQzrWoChkJ5moqqqKclTxw+PxAPDhhx9qHIkQQgghBiOJuYgL7e3tdLR3YMq48MLPXtlJJqqrKqMYlRBCCCFE5EhiLuLCcBZ+9sq2m2huaaWnpydaYQkhhBBCRIwk5iIunDp1CgDzEFol9sqRBaBCCCGEiCOSmIu4UFpaijnVjsFqGfJrsu3hlomyAFQIIYQQ8UAScxEXSkpLMWYNvYwFYEySCaOiSGIuhBBCiLggibnQvfb2dlpbWjBnpQ/rdUZFYZzdTGWlLAAVQgghhP5JYi50r7i4GADLmLRhv3a83Ujl6YWjQgghhBB6Jom50L3i4mIUgwHzMEtZAMYnm2luacHtdkchMiGEEEKIyJHEXOheUVER5kwHitE47NdOSA53ZpFyFiGEEELonSTmQtcCgQAnT57ENIIyFvgkMS8vL49kWEIIIYQQESeJudC1yspK/H4/ljHpI3p9usVIqtVEaWlpZAMTQgghhIgwScyFrvUm1OYRzpgrisLkFBPFRUWRDEsIIYQQIuIkMRe6VlpaitFmxZiSNOJzTEm1UFdfT1dXVwQjE0IIIYSILEnMha4Vl5RgynKgKMqIzzE5NbxbaElJSaTCEkIIIYSIOEnMhW719PRQU1MzojaJ/U1KsaAokpgLIYQQQt8kMRe6VVZWBqqKOWtk9eW9bCYDOXaLJOZCCCGE0DVJzIVu9S38HGViDjA51UxpSTGqqo76XEIIIYQQ0SCJudCtkydPYkpJwphkHfW5pqRa6Hb3UFtbG4HIhBBCCCEiTxJzoVtFRUWYMkdXX95LFoAKIYQQQu8kMRe61NzcTEtLC5ZxGRE5X7bdhM1kpLi4OCLnE0IIIYSINEnMhS71JtDmsZFJzA2y0ZAQQgghdE4Sc6FLRUVFGExGzJmpETvnFIeFyqoq3G53xM4phBBCCBEpkpgLXTp+/DimMWkohsj9iE5zWFFVlaKLfNY8FAppHYIQQgghBiGJudCdzs5OKioqsGRnRfS8kx0WDIrCiRMnInreeNHZ2QmEN24SQgghhP5IYi50p7CwEABrTmZEz2s1GshLtVBYeDyi540XwWAQgEAgoHEkQgghhBiMJOZCd44dO4bBZMI8ZvQbC51pmsPCydKTeDyeiJ9bCCGEEGI0JDEXunPo8GHM49IjWl/ey5luJRgK9c3KCyGEEELohSTmQlcaGhpoqK/HOmFMVM4/Nc2K2aBw5MiRqJxfCCGEEGKkJDEXunL48GGAqCXmZoNCfpqVI6fHEUIIIYTQC0nMha4cPnwYU4odoyM5amPMSrdSXVNDU1NT1MbQI1VVtQ5BCCGEEOchibnQDb/fz5EjR7CMz0JRlKiNMyfTBsCBAweiNoYQQgghxHBJYi5048SJE3i9Xqy5Y6M6TrbdzFi7mX379kV1HCGEEEKI4ZDEXOjG/v37UYxGrDmR3VhoMHMyrBw/dkzaJgohhBBCNyQxF7qgqir79u/HkpOJYjJGfby5mTYCwaB0ZxFCCCGEbkhiLnShurqapsbGqJex9Mp3WEkyG6WcRQghhBC6IYm50IX9+/cDYJs4LibjGQ0Kc9KtHNi/j1AoFJMxhRBCCCHORxJzoQv79+/HkunAmGyL2Zjzsmx0dnVTVFQUszG1JO0ShRBCCH2TxFxorrOzk+LiYiwxKmPpNSvDhsmgsHfv3piOK4QQQggxGEnMheaOHj2KqqpYJ8Q2MbeZDMxMs7J3z+6LYjb5YvgahRBCiHgmibnQ3NGjRzFYzJiz0mI+9rwsGw2NTVRWVsZ8bCGEEEKI/iQxF5pSVZVDhw9jzs5EMURvt89zmZuVhALs2bMn5mPHmsyYCyGEEPomibnQVH19PS3NzVjHR39TocGkWoxMcVjZexEk5kIIIYTQN0nMhaaOHz8OgDUnU7MY5mXZKK+ooLGxUbMYhBBCCCEkMReaOnHiBKYkK0ZHsmYxzMsKt2iUzYaEEEIIoSVJzIVmVFXl2PHjmMZloCixry/vNTbJTLbdLOUsQgghhNCUJOZCM42NjbS2tGDJztA6FOZl2TjhctHV1aV1KFGj5cWPEEIIIS5MEnOhmWPHjgFgydauvrzX3EwboVCIgwcPah1K1EhXFiGEEELfJDEXmjl69CimJBumtBStQyEv1YLDaroo2iYKIYQQQp8kMReaCIVCHDl6FFOOtvXlvQyKwtwMK4cPHcLn82kdTlTIjLkQQgihb5KYC01UVVXR1dmpWf/ywcwbk4TX5+Po0aNahxIVergAEkIIIcS5SWIuNNFby23NGaNtIP3MSLOSZDJKOYsQQgghNCGJudDEgQMHMGc6MCbbtA6lj8mgUJBpZe/ePQSDQa3DibhQKKR1CHFj+/btvPLKK1qHIYQQ4iIjibmIua6uLoqKirBOHKt1KGeZn5VEd7ebwsJCrUOJOKkxH7rf/e53rFu3TuswhBBCXGQkMRcxd+jQIVRV1WViPivDitloSMhdQHtnzKXWXAghhNAnScxFzB04cACjzYo5K03rUM5iMRpwplnYu2dPws0wJ9rXI4QQQiQaScxFTAWDQQ4ePIhlQhaKQZ8zt3OzkmhuaaG8vFzrUCJKasyFEEIIfTNpHcBwOJ3O+4F/nPHwGy6X606n0zkZ+BNwBVAB/KvL5Xon1jGK8ysuLsbtdpOeO1PrUM6pINOGAuzbt48pU6ZoHU7EyIy5EEIIoW/xNmNeAKwDxvf7+LzT6VSAN4BmYDHwV+A1p9M5VatAxeAOHjyIYlB01b/8TKkWI1McVvbt3at1KBElibkQQgihb3E1Yw7MAQ67XK66/g86nc7rACewwuVydQLHnU7nDcAXge/HPkxxLgcOHsA8NgODxax1KOdVkGnlrbJympubycrS70XEcPS2gJQEXQghhNCneJsxnwO4Bnl8GXDgdFLeaxuwPCZRiSFpbW2lsqIS6wT9bCp0LvOykgDYv3+/xpFEjtSYCyGEEPoWNzPmTqfTAkwDbnM6nT8CFOAV4D8Jl7TUnPGSeiA3pkGK8zp8+DBAXCTm45JMjLWb2bd3LytXrtQ6nIhIxE2ThBBCiEQSN4k5MINwvN3A3YST9N8AqYAN8J5xvBewxjJAcX6HDh3CZLdhykjVOpQLUhSFeZlWNh8/jtvtxm63ax3SqPWWsEgpixBCCKFPcVPK4nK5jgFjXC7Xl10u1yGXy7UW+P+ARwAPZyfhVsAd2yjFuYRCIQ4fOYI5JzNuNriZl5VEMBjkwIEDWocSEVLKIoQQQuhb3CTmAC6Xq/mMhwoBM+EylpwznssBamMRl7iwsrIy3N3dWMfrv4yl1+RUCw6rid27d2sdSkRIYi6EEELoW9yUsjidzlXAH4BJLpfLd/rhS4A2YCfwPafTmexyubpPP3fl6ceFDhw9ehSAblcF7pJKkgvysU0cC4C/pYOOvYV9x6avWIgxKXwDxF1aRU9pNQAGm5WMqxb2Hde+p5BAawf+1q6oxGw4Xc6y59BBvF4vVmt8V0ZJYi6EEELoWzzNmG8mvODzj06nc6bT6bwV+O/THx8B5cBzTqezwOl0fpdwp5Y/aRWsGOjIkSMoJiOKMT7KWHrNz0rC5/P3LVyNZ5KYCyGEEPoWNzPmLper2el03gT8EtgPtANPAU+4XC7V6XR+CngW2AeUAne5XK4yreIVn/D7/biKirBPz8WxePZZz5szHWTduHTQ19qn5WKfNnhznbTT52r5cB/eqsbIBdzP9DQrdrOR3bt3s3jx4qiMESuBQEDrEIQQQghxHnGTmAO4XK4DwLXneK4EuDq2EYmhKCkpIeD3k5oT3Y16nj7WhEGBG3IdzM60AVDV5WPdyba+Yx6alYXDYgRgV303u+vDlU8Os5GHZn8S39rSNqq7wxVTmVYDB/bvIxAIYDLF1a/MADJjLoQQQuhbPJWyiDhVWBiuH7eMy9A4kpEZbzfj7vFw7NgxrUMZFZkxF0IIIfQtfqf/RNw4dvw4lkwHBqs5quN8pWAMNtPAa83cFAvfnD9u0OOXZiezNDt50OdWTUvv+7cvqHKwycOhQ4dYsGBBxOKNNUnMhRBCCH2TGXMRVYFAgJLiYkxxOlsOYDEqTEuzcOjgQa1DGRW/3691CEIIIYQ4D0nMRVSVlZXh9/vjtoyl1+wMK7V1dTQ0NGgdyohJYi6EEELomyTmIqqKiooAsIxN1zaQUZqdEV5MGs9tEyUxF5Hm9/t5//335WdLCCEiRBJzEVVFRUWYU+wY7TatQxmVsUkm0qwmjh8/rnUoIybJk4i0LVu28Je//IUtW7ZoHYoQQiQEScxFVJWUlmLMcmgdxqgpisI0h5kThcdRVVXrcEZEEnMRab0XqvF8wSqEEHoiibmIms7OTlqamzEnQGIOMC3NSlt7B3V1dVqHMiKSmItIa21tHfBfIYQQoyOJuYia8vJyAMwZiZGYT0+zAnDixAmNIxkZaZcohBBC6Jsk5iJqTp06BYA5MzES83FJJlItpoRIzOO1HEcIIYRIZJKYi6iprq7GlGTDYLNoHUpEKIrC1FQzJ07vZBpvfD5f37+lrEUIIYTQH0nMRdRU19RgcNi1DiOi8tMsNDY10dLSonUow9Y/Me/p6dEwEiGEEEIMRhJzERWqqlJTU40pbfAt7+PVNEe4ztzlcmkcyfD1T8w9Ho+GkYhEoSiK1iEIIURCkcRcREVHRwc97h5MjsRKzCckmzEbDJw8eVLrUIbN3W+W3O12axiJSDSSoAshRGRIYi6ioqamBiDhZsyNBoWJKWZKSkq0DmXYvP1mybu6ujSMRAghhBCDkcRcRMUniXmKxpFEXl6KmVOnThIMBrUOZVj6z5h3d3drGIlIFNLdZ+gaGxvlglgIcUGSmIuoqKmpwWAyYbDbtA4l4vJSLfh8fqqrq7UOZVj8/WrMOzs7NYxEiIvPd7/7XX79619rHYYQQuckMRdRUVUdXviZiLWnE5LNAFRVVWkcydCdublQR0eHRpEIcXHyeDwcP35c6zCEEDonibmIiqqqKgypidUqsde4JBMGRYmrxPzMxZ7t7e0aRSKEEEKIc5HEXEScx+OhtaUFU3ri1ZcDmAwKY+1mKisrtQ5lyM7sWy6JuRBCCKE/kpiLiEvUjiz95SQZqa6Kn8T8zMWe8bhBkhBCCJHoJDEXEZfIHVl6jbGZaGxqJhQKaR3KkJy52LOlqUmjSIQQQghxLpKYi4irqakBRcGUkpg15gBjbEaCwWDczDz3nzG3GhXaOjrOWhAqxHD1tktMxEXeQgihBUnMRcTV19djTrGjGBP3xyvLZgKgoaFB40iGpv+MeYbViKqqNDc3axiREEIIIc6UuJmT0ExtbS2G1CStw4iqjNOJeVOclIT0b4845nTs9fX1WoUjhBBCiEFIYi4iSlVV6urrMSZoq8ReKebwr068bNTT0dEBhnC5wZgkScyHKl7WEAghhEgMkpiLiPJ4PHh6ejAmJ/aMuc2oYFCUuErMDTYLAA6LEYvRQF1dncZR6Z/f79c6BCGEEBcRScxFRLW1tQFgSLJoG0iUKYpCstlIV1eX1qEMSVtbGwabFQAFyLab4qoPu1YkMRdCCBFLkpiLiOrduMZ4OglMZEkmw1k7aupVa1sbBqu57/PxdhOVFRUaRqRfwWCw79+SmJ9fb1cWIYQQkSGJuYio3sTckJT4ibnJEB+Jm6qqtLe1Dfh/Mt5upr2jY8CiUBHW//+pz+fTMBL9kzaJQggRWZKYi4jq3fpdMZs0jiT6jApx0Qu8q6sLv9+P0W7re2xicnj2vKysTKOo9Kt/Mi6J+fnJjLkQQkSWJOYionpnGxO5h3kvU5wk5r39yo32T2bMc1PCawBOnTqlSUx61n/G3Ov1ahiJEEKIi03iZ08ipnoTVcWQ+Le4TYpCIA5KWXp7rRvtn3TKsZsNjEkyc/LkSa3C0q3+yXjvHSAhhBAiFiQxFxHVN4NsSPwfLUWJjz7Xvf3KjWds+jQpxcTJkhItQtI1j8fT929JzM9PSlmEECKyEj97EjHV+0Z9MSwKM8RJYl5XV4fRakGxmAc8PiXVQnNrKy0tLRpFpk/9k/Hu7m4NIxFCCHGxkcRcRFRvQn4xzKQZFCUuEvPq6upBd2Kd4gjXnBcXF8c6JF3rn4zHywZSWrmYLsSFECIWJDEXEWXoLWFJ/LwcBQiFghc8TkuqqlJeUYEpI+Ws5yYmmzEbFFwulwaR6Vf/3vSSmJ9f74XpxXAhLoQQsZD4Pe1ETPUm5i2b9p61ANSU4SBt8exBX+curaKntHrQ5xyXzcac6Rj0ueYNu/C3arj7ps5nCltaWuhxu3FkTDnrOZNBYarDwtEjh2MfmI713821tbVVw0iEEEJcbCQxFxGlZWnH08eaOLMZzMRkC6umpQ96/K76bnbXD15DfFd+el9LwTM9ebgBAKvRgIK+E/PedojmjNRBn3em21hfVkNraysZGRmxDE23+ifmva0mxeBkplwIISJLEnMRUb3bmWfdsBjFZBzy6+zTcrFPyx32eFk3LqXlw314qxqH/dpRU/VfW1tcXIxiMGDOchD0nt3acWaGFcrgyJEjXHXVVbEPUIf6l680n241KQbXm5hLgi6EEJEhibmIqN7E/Kyp6xj4SsEYbKahL5tYmp3M0uzkYY/zzfnjAHjqaOMnNfU6VVxcjDnTgWI0Amcn5hOTzTisJg4cOCCJ+Wn9Z8xbWlsJBAKYTPKncjC9v++SmAshRGToO6sQcacvMdf5THIkBEPoOmELBAKUlpZiGpN2zmMMikJBhpVDBw8O2PHyYtbRb8Y8FAr1bdAkztabkMdDdyIhhIgHkpiLiAoGgygGRfclHpEQVFWMOk7My8rK8Pv9WMadv3Z8bqYNj9fL8ePHYxSZvrW3tw/4vK6uTqNI9K83IZfEXAghIkMScxFRoVAIlIvjxyqo6nvG/MSJEwAXTMxnpNuwmgzs3r07FmHpXltbGwA2Y/jisrp68G5B4pM7ZHK3RQghIuPiyKBEzPTOmF8MfCpYLIN3btEDl8uFOTUZY5L1vMdZjApzM2zs3rWLQCAQo+j0yev14jm982ey2YDDaqK8vFzjqPRLEnMhhIgsScxFRAWDwYuijAXAE1Cx28/eUVMPVFXFVVSEaey568v7WzQuiW63m8OHL+6e5i0tLQM+z7WbKDvdclKcrTchl8RcCCEiQxJzEVFGoxE1dHF0aOgJhEhKStI6jEHV1dXR1dl5wTKWXs50G3azkW3btkU5Mn1rbDzddtMY/tOYm2KmprYWj8ejYVT61ZuQe71ejSMRicLn81FZWal1GEJoRhJzEVE2m41QIJDw7dNCqoo3ECQ5efjtFmOhsLAQuHB9eS+TQeHSsUns3bNnQLvAi019fT0AhtM9+PPTrIRCIVwul5Zh6ZYk5iLSfv3rX/Pd7363b62HEBcbScxFRFmt4XpmNRDUOJLo8gRUVNBtKcuxY8cwJdkwOoZ+4bA0204gGGT79u1RjEzfGhoaUIwGMIYT86kOC0ZFkY415yCJuYi0Q4cOAdDa2qpxJEJoQxJzEVE2mw1I/MS8+/TXl5o6+Fb3WlJVlWPHj2HKzhhWvX9uioXcFAsfbtqU8Hc8zqWmpgaTIxlOf9usRgOTUy0cPXJE28CEEEJcFCQxFxHVW3Ot+hO7u0e3P9y3WY+JeXl5OR3tHVjHZw37tcuy7VRUVnLy5MkoRKZ/VdXVZ91lmJNp5VRZmWw0JIQQIuokMRcR1ZuYhy6SxDwlJUXjSM528OBBAKwTxw77tZeOs2MxGti0aVOEo9I/n89HU2NjeMa8n/lZ4Z/pPXv2aBFW3LhY77IIIUQkSWIuIqq35lr1JXZi3uELl7Kkp6drG8gg9h84gCUr7YL9yweTZDKwcIyNHdu343a7oxCdflVVVaGqKqaMgXdBxtnNjE+2sGvXTo0i06czE/Hu7m6NIhFCiMQhibmIqN7EPORL7L7Grd5wv3a9Jebt7e2UlJRgGcFsea8rxqfg9fnYsmVLBCPTv96NhMwZZ5cnXTrWRlFRMbW1tbEOS7fOXPB5Zg94IYQQwyeJuYiovhnzBC9lafMFSXc4MJlMWocywN69e0FVseVlj/gck1MtTHZYee/ddwmFQhGMTt/KysowmE0YU8/utLN4XDKKAps3b9YgMn06c4a8rwe8EBFwsWxUN1KqqvLss89SVlamdSgiwiQxFxF1sSz+bPMGyRwzRuswzrJ7927MjmRM6aOrfb96fDL1DQ199eoXA5fLhSkrbdCEIM1qZHaGjS2bNxMIJPbP9lB1dnYO+LyhoUGjSIS4+LS1tbFx40b+9Kc/aR2KiDBJzEVE9bZLTPTFn83eEOPGjdM6jAHa2to4euwYlrzsUc82LRiTRJrVxD//+c8IRadvbrebispKLOPSz3nMleOTaWtv5+OPP45dYDrW0dEx4POqqiqNIhHi4iX93hOPJOYiokwmE2azOaFnzAMhlZYeP+PHj9c6lAF27NiBGgqRlD9h1OcyGhSumZBMYWEhxcXFEYhO31wuF6gqlnGZ5zxmVoaNnGQLb61fLx1IGCQxl23URQRJKYu4WEliLiLOarMl9AZDTT0BVCAnJ0frUPqoqsrmLVuwZKVhTotMC8fLxydjNxt58803I3I+PTt8+DCK0YhlbPo5jzEoCtdOTKayqqpvd8KLWf+ZuvF2E5WVFQSDift7L4QQsSCJuYg4m82G6k/cN+iGnvDdAD0l5idPnqSyogLbtIkRO6fVaGDFeDv79u1L6DIFVVXZt38/lpxMFJPxvMdeOtZOps3MmjUvXVQLYwfT3Nzc9+8pDiser49KmTUflNxhEUIMlSTmIuJsNhuhBF4g1+oNf216qjHfuHEjBpORpKmRLa9ZMSEFi9HA+vXrI3pePampqaGpsRFr7oVbTJoMCrdMTqW8vIIdO3bEIDr96p+YT3VYgNMlQeIssmBYCDFUkpiLiEtK8BnzNm8Qs9lEaurZ/a614Ha72b5jB9Yp4zFYzBE9d4rZyLJsOzu2bx+QiCWS3bt3A2DLHdqF1qKxSeSmWHh5zRp8Pl80Q9O1mtpaMIbfQjKsRjJsJo4fP65xVPrk9yf2vg7RYDBIeiIuTvKTLyLObrdDAs8QtXqDZGVk6mZx0scff4zf58M+PTcq579mYgqqqvL2229H5fxa+3jnTixjMzDabUM63qAo3DHFQVNzM++8806Uo9OnUChEQ0M9xuSkvsec6VaOHjkis8OD6J+YS1mLEOJ8JDEXEWe321F9ifvm3O7TVw/zTR9uwpyeinlMWlTOn2kzsWhsEps2bqSrqysqY2ilpqaGqspKbJOHtyHTzAwb87KSeH3duouyXVljYyPBQBBj8icXM3MybPR4PBQVFWkYmT71v7Mis+fn13vhopeJDyFiTRJzEXF2u51QAifmbT6VzMxzt9WLpYqKCk6dPEXS9IlRfSO7PjcVr8/Hhg0bojaGFnrrxG2Th7+Q91NT0wgG/Lz44ouRDkv3KioqADCmfDJjPjPdismgsG/fPq3C0i2v19v3756eHg0jEULonSTmIuKSk5MJ+XwJecs2pKp0eANkZGRoHQoAH330EYrBEJHe5eczPtlMQaaN9959Z0CSEc9UVWXL1q1Yx2cNuYylvzFJJq6dmMK2bdsoLCyMQoT6VV5eDoqCMcXe95jNZKAgw8a2rVtkVvgM/X9n3G63hpHED5kxF5Gya9cuXn/9da3DGDJJzEXEpaamoobUhNxkyO0PEVT1MWPu8/nCiWXeOAxWS9THuz43lc6ubjZt2hT1sWKhqKiIpsZGbFNHflGzclIqWUlmnn3mTxdVbXVFRQVmRzKKYWDytGx8Mp1d3TJrfgaPx9P37/b2dg0jiR+JOLEjtPGb3/yGl19+WeswhkwScxFxaWnhWudQT2LMrPbX5gt3m9HDjPnevXtxd3djnz4pJuPlp1mZnmZl/ZtvJkQ3ko0bN2Iwm7DlDa++vD+L0cDd+Q5qaut46623IhidvpWeLMWYcXZXIme6lQybiQ/ef1+DqPSrf2J+5o6pYnCSmIuLlSTmIuJ6E/OgJ/6TtzO1eMKJ+RgdLP58//33MacmY8mJ3ez9TXmptLW389FHH8VszGjo6Ojg450fY8ufgMFsGtW55mQmsXBMEmtfe43q6uoIRahf7e3ttLa0Ys5ynPWcQVG4anwyxwsLpad5P/3ryi/GxcJCiKGTxFxEXN+MeQIm5k2ecLlCdvbIZ1kjoaqqCpfLhW1GbkxrMaenWcl3WHnzjTfiunTjww8/JBgIYnfmReR8d09Lx2qAp576Q8LvCFpaWgqAOWvwLkCXj08mxWLktddei2VYutbT00Pvb2lTU5OmscQLmTE/v961CrKYOPFIYi4iLpFLWZo9AZLtdpKTkzWN44MPPkAxGrBPmxjTcRVFYWVeKi2trWzdujWmY0eKz+fjnXffwTo+C3NaSkTOmWoxsio/jdLSkwnb772Xy+VCMShYzpGYW40GrpuYwtGjRzlx4kSMo9On/ol5Q0ODprGIxNBbTpgIZYViIEnMRcSlpqaiKEpizpj3BDSfLfd4PGzZsgVbXg4GW/QXfZ5pVrqVSakW3nzjjbicHf7oo4/oaO8geW5+RM+7aGwS87KSeHnNmr52gomosLAQc1Yaisl4zmOuGJ+Mw2rixX/8Q2Y+ge7u7r5/19fVahhJ/JCfG3GxksRcRJzBYCA5JYVgAs6YN3hCTJgY21nqM+3YsQOPx4N9ZmwWfZ5JURRuyE2lvqGBnTt3ahLDSAUCAd54800sYzOwZEe2Nl9RFD4zPR27SeHJ3/42YdpK9ufxeDh58iTmcedf/Gw1GrglL5XikhJ2794do+j0q//GXLW1dXFdBhYr8XjRL0QkSGIuoiItLS3hZsy9wRCtHj8TJkS3Z/j5qKrKhg0bMGekYh6brlkc87Js5CRbeH3d2rh6A924cSOtLS2kzM+PSm1+isXI6hnpVNfU8MILL0T8/Fo7fPgwoVAI64QLL35ekm1nfLKFF//x94v+dntnZ2ffv/2BADU1NRpGEx9kxvz8pM974pLEXETFmKwsVLfnwgfGkXp3eJZLy8S8qKiIiooKkmZO0vQPs0FRWJmbQlV1DXv37tUsjuHweDysXbsWa3YmlvHR66rjzLBxXW4KGzduTLh+3vv27cNotWC5wIw5hH9G7prqoKGx6aJfCNra1gbAxGQzAGVlZdoFEyckMRcXK0nMRVRkZ2fja+2k6b1dNG/Yhae6se85f0sHzRt29X30L3lxl1b1Pd665eCAc7bvKex7ruvoyVh9KX3q3eHdDCdqWMqyYcMGDBYzSaPYFCdSLhmbxDi7mbWvvRoXs+Zvv/02nZ2dpFwyM+oXNbdMTiM3xcIfn36KttNJWbwLBALs278fy4QxKIahvXXMzLCxLCeZt956i+Li4ihHqF+9LRLHJpmwmQwUFRVpHJH+xcPfFCGiQRJzERXjxo0DVQ1/JIiabj9ms4mcnBxNxm9tbWXXrl0kRaD3diQYFIUbclOoqKzSfR1xW1sbb65/E1teNpYYlACZDAqfc2bgcbt56qmnEmL279ChQ7i7u7FNHt7P/51T00i3Gnn6qT9clCUtoVCI9vY2AAwKzEizcOjgwYT4mYgmSczFxUr7d3eRkMaNGweA47LZWMYMbKtmznSQdePSQV9nn5aLfVruoM+lLZ593jGfPtaEQYEbch3MzrQBUNXlY93Jtr5jHpqVhcMS7iaxq76b3fXhbgkOs5GHZmf1Hbe2tI3q7nASMTvDxg2THFR3+5mUOwmj8dzdKKLp3XffJaSq2GdN1mT8wVw2zs6m6m5eXrOGxYsXa/a9uZCXX34Zvz/AmEtmxmzMHLuZT01N49XDh3n77be59dZbYzZ2NGzduhWjzYp14vDKgGwmA/dNT+cPR+t47rnn+PKXvxylCPWpvb2dYCDYNws2O8PGkZIWqquryc0d/G+dkMRcXLxkxlxERW+5R6C18wJHxgdVVanuDjB5yhRNxne73bz//vvY8rIxpdo1iWEwBkXh1smp1NXXs2XLFq3DGVRFRQWbN2/G7pyEyRHb/vNXjE9mXlYSL730EiUlJTEdO5K6urrYv38/1ik5Qy5j6c+ZYWPlpFQ++uijuN81drgaGxsHfD47IzxpsH//fi3CiRvSuUZcrGTGXERFdnY2tqQk/C3tQGxmhb5SMAabaWDSkJti4Zvzxw16/NLsZJZmD56orZqWPuDzVk+Abn+QyZO1ma3euHEjHo+HrIKpmox/PnMzbUx2WFm3di1XX301hhEkbtH097//HYPFTMq8aTEfW1EU7p+Rwc8PNvDkb3/D40/8VPPNqUbio48+IhAIkD595L/LN092UNbp4y9//jNTp07V7Hcp1urr6wd8nmEzke+w8tGHH3L77bdLd41++s+SB4NBDSMRQjv6egcVCcNgMJA/dSqB5g6tQ4mIU53hspbp06fHfGyv18tbb72FdXzWOXdb1JKiKFw3MYWm5mYOHDigdTgDHD16lCNHjmCfOxWDNfabMQHYzQYedGbQ3NzCs88+G3e1xaFQiA3vv49lXAbmjNQRn8egKDzozMRuhF/+4hcDensnstraWlCUcIH5aUuz7dTV18si0DP0bjMP4Pf7NYxECO1IYi6iZtq0afhbO1GD8V8rWNbhw2Ixk5eXF/OxN23aFO4mosGM71DNzbKRbjPx3nvvaR1Kn1AoxD/+8Q9MyUkkO2P//62/KQ4rN09OZefOnezYsUPTWIbr4MGDNDU2Yo/A9zDVYuQLszJoaWnmySefvCjqiGtqajCfUX62cGwSVpOBjRs3ahSVPnV0fDKRI6Us5yd3WhKXJOYiaqZPn44aCuFvbtc6lFE71elj2rTpmEyxrf4KBAKsf2s9luzMiO1Uua22iycPN1DY8kmf+aouH08ebuj76PB9cht5V3133+N/LWwecK61pW08ebiB3x9pJNtm5OjRo+EZQh3Ys2cPZWVlJC+YjqKDRanX56Yy1WHlL39+lqamJq3DGRJVVVm7bh2mlCRsedkROecUh5VP56dx5MgR1qxZE5Fz6ll5RTmGM9Y2WI0Glo6zs2PHdurq6jSKTH/6J+aJuHNuJMXbnTcxdJKYi6iZOTPcAcPX2KpxJKPjDYao7vL1fT2xtH37dtpa20iZq7/a8jPlpoQ3T9mzZ4/GkYRny1997TXMaSm66PkO4VKO1TMzCPh8/PHpp+PijfXo0aOcLC3FXjB1RIs+z2X5+BQuz0lm/fr1ul00HAlut5v6unrMWY6znrthUipGRbnoN1/qr39ifjG21hwOva3lEZEjiz9F1KSlpZGdk01bQysUaB3NyJV1+AipMGvWrJiOq6oq6996C3NGakR3qrxyfArX5g6sFY7UItmidh979+zhjjvuiEisI7V7926qq6pIv3I+ikE/t3zHJJm4fYqD144dY/v27Vx55ZVah3ROqqry6muvYbLbztnCdDRWTUun0RPgT3/6E1lZWRQUxPEfiXMoLy8HwDzI2hCHxciK8XY+3LGdT33qU9I6kU82YgKZMRcXL7nkElE1e9ZsAo3tcTE7eC6l7V4URYn5jPnhw4epqa7GPntK3NQTzsu0UVJaOuANNtZCoRCvvvoq5rQUbJPHaxbHuVwxPpm8VCsvPP+8rhdA7t69m+KiIuzz8lGMkX+rMBkUvjA7i7E2I7/65S+orq6O+Bha613cOVhiDnBdbio2o5Hnn38+rv9GRkpLS0vfv/svBBXiYiKJuYiqWbNmEfT6CLTpNwG5kNIOH1OnTCEpKSmm436wcSNGm5WkKfpLLs9lflb4e6TlTqA7duygpqYmXFuuo9nyXgZF4d7paXR1dfHSSy9pHc6gfD4fL/z975gzUrFPnxS1cewmA4/MycQYDPDTJ56gubn5wi+KI8ePH8ecnorRNnhHoBSzkX/JS+HIkSPs3bs3xtHpT/8L+u7ubg0jiR9yQZd4JDEXUTV7dni3Tl9dfL7h+oIq5Z0+Zs0+/66jkdba2sr+/fuxTZsQldnKaMlJNjM+2cLOjz/WZPxAIMArr76KOdMRscWK0ZCbYmHF+GQ+/HATZWVlWodzlrfffpvmpiZSL50V9YubLJuJLxdk0t3RxhNPPD6gzjieBQIBTrhcmLMzznvclRNSGJ9s4fm//e2ir6vuvxmTnu8m6YEk5Ikrft7xRVwaO3YsWWPG4K2PzwWgZZ1eAiE15vWvH374IWoohH0UG7po5ZIxNlxFRZrMfm7dupXGhgZS5k8fdflPuzfY140mkh1snjzcwAeVHdw02UGy2chf//qcrt5km5qaWLduHba8bKzjs2Iy5qQUC1+anUVDXR0/+6//oqenJybjRlNhYSF+nw/rBdaHGBWFVfkOmpqbefHFF2MUnT419EvM29vjv5tXNF0MrUYvVpKYi6grmDOHQEOrrpKPoSpu82IwGHA6nTEbMxQKsenDD7HmZMV8C/lIuGRsuGfzxzGeNQ8EAqxdtw5LVhrW3LExHXsk7CYDt+al4nIVaVr6c6bnn3+eoBrCcVlsFztPT7fy+VmZlJWV8dvf/CbuE4+9e/diMBmHdHEzI93GVRNSeO+999i3b18MotOfQCBAe1tbeDMmoKUlPu+yxkq8/36Ic5OuLCLqCgoK2LJlC4HWTsyZZ7cN07Pidh/5U6dit9svfHCEHDlyhJbmZoxpyTRv2EVyQT62ieFE09/SQcfewr5j01csxJhkBcBdWkVPaXgBncFmJeOqhX3Hte8pxN/UFpP4xyaZmJxqYdvWLdx2220xGRPCs+XNTU1kXLsoIotl06zGQTvVRKqDDcDSnGQ+rOlm/ZtvsmTJEs0X+R45coQ9e/aQsnAGxuTYrqkAmJuVxN3T0njl8GFeffVV7r333pjHEAmhUIg9e/dgmTAGxTS0Hvp3TE2jtMPH0089xX/97GdkZJy/BCbRNDc3EwqFMNgshDw+mptbLvyii1g8TnSJoZEZcxF1c+fOBcBbG18zIJ5AiIpOL3M0KGNBUTTbQj4SLhtnp6Kyqq9dXLSFQiHWvf56eLZ8ov5ny3sZFIWrJyRz8tQpXC6XprEEAgGe++tfMaUmkzJHu775V4xPYVlOMq+//roueuKPRGFhIW2tbdjycob8GpNB4UFnBj5PD7/59a8vui3pGxoaAFCs4f0QAoGA1JmfRzAYvPBBIi5JYi6iLiMjgwkTJuCLs8T8ZIeXkPrJhUUstLe3s2/fPpJnTWbMTUvJunFp32w5gDnTQdaNS/s+emfLAezTcvse7z9bDpC2eDbpV18Sqy+DS8YmYTQoMdt+fu/eveFt4wumaj7rPFyLx9mxm4288847msbx/vvvU1tTQ+qlTs0XHN89LZ28VAt/+P3vdbOT7HBs3rwZg8WMbdLgd1bOJdtu5v4Z6RQVF/PHP/7xopoVra+vB8Bg+WRCoqqqSqtwdE8S88QlibmIifnz5+NraEWNoz8mxW1eTEZjTPuXb9u2jWAwSFIcLvrsL8VsZHKqhWNHj8ZkvLfffhtTih3bJP12YjkXi9HApWNsHDp4ULNZ0o6ODl597VWs48fooj7fbFD4wuxMDKEAv/uf/yEQCGgd0pC53W527d6NbXL2kMtY+rtkrJ1bJjvYvn0769ati0KE+lRfX49iNGKwfFJhW1lZqWFE+hZPvxNieCQxFzExf/581GAQb1381A0Wt/uYMWMGFktsSkpUVeXDDz/EMjYdc3pKTMaMpnyHhfLycjwez4UPHoVTp05RVFREkjNPl33Lh8KZYcPn91NcXKzJ+C+++CIej4fUxbN0c8chw2riM9PTOXnqFK+++qrW4QzZli1b8Pt8JM0Yef/3lZNSWTzOzquvvsqWLVsiGJ1+1dbWYkq1w+kfP5NBobS0VNugdEwS88Qliz9FTMyZMweL1YK3qmFAaYZedfuDVHf5uDuGZSynTp2ipqaGtKWJsTV5vsPCB5WdlJaWRrXd5KZNm1CMRuzTJ0ZtjGibnmbFoIQXX86ZMyemYxcVFbF582aS50zFnDayC8J1J9tIMhm4IdfB7EwbEG4rue5kW98xD83KwmEJzyDvqu9md314AxmH2chDsz/pXLK2tI3q7nA/79kZNpbnJLN+/ZvMmzcv5m1LhysUCvHue+9hGZuO5Ry7fQ6Foih8ZkYG7b4gTz/9NDabjSVLlkQwUv2pravFkJoEp8t3pjksHD92TOOo9OtiW4NwMZEZcxETFouF+fPm46tuiou6yaI2Lyowb968mI25detWFKMB2+ShLxjTs0kp4TsN0awT9Xg8bN++HdvkbAwWc9TGiTabycDEFAslJSUxHTcUCvHnv/wFk91GyvxpMR17qO7MT2NMkpmnn35K9/3NDx8+TEN9PXZn3qjPZTIofHFOFlNSLTz55JMcPHhw9AHqVCgUoqGhYUB72OlpVpqamwdsOiQ+cbFvRpXIZMZcxMyiRYvYu3cvgZYOzKOYTYoFV5sXe5KN/Pz8mIwXCATYtn071oljMVjjN8HsL8VswGxQaGpqitoYu3fvxuPxkBnnNfkAE5PNHC0rQ1XVmJWTbNy4kYryctKvXIDBPPK3g7vy05mRbhvwWCTbSt4/PZ0nDzfyj3/8gy9+8YsjjjPa3njjDUzJScPqxnI+VqOBRwqy+P2RJn71y1/y3ccei/kdlVhoamoiGAhiSrUTaA93YpmRboVyOHjwICtXrtQ4Qv2RxDxxyYy5iJlLL70Uo9FIT5m+uyyoqoqrzUvB3HkYjcNfvDUShw4doruri6Rp8VuOcSZFUciwmaM647Vr1y4wGOg8WEzzhl14qj8Zy9/SQfOGXX0fwR5v33Pu0qq+x1u3HBxwzvY9hYT6HRsrE5PNdHV309ISm3UY7e3tvLTmJaw5Wdim6PsuTX6alasnprBx40aO6bS8weVy4XK5sM+eHNGuNnaTga/OzSLLauC/f/YzzdtqRkNdXR0Axn4z5jl2MznJFrZv26ZVWLrm9X7yN0rqzYcmXjZlksRcxExqaioLL7kE76k6VB3/gtS7A7R6AsyfPz9mY27ZsgVjkhXrhPNv3x1vMq0KDafboEVaT08PR44cwWA1o5P1iqOSYw/PWNfU1MRkvDVr1oQXfC6ZrZsFn+dzy+Q00m0m3nj9da1DGdTrr7+O0WaJSkelFLORr83NwmFS+a+f/jTmJU/R1tsS88ydji8da6OouFjKWQbhdrv7/t3e3q5hJPEjXuryJTEXMXXVihUEejy63mzoSEu4jnXRokUxGa+rq4v9B/Zjm5yDYkisX8ksm6lv45BIO3ToEIFAgPQVC/r6t0eq57uh37GxkmULJ+bRLP3pVVNTw+bNm0mamTfiBZ+xZjEqXJ5t5+ixY7rrbV5cXMyhQ4ewz5o8qpKg80mzGPnG3CySDSF++sQTnDp1KirjaKGurg6D2YTBNrAD1qVj7SiES67EQP3XW7S2tmoYSfxIuMTc6XQ+7HQ69d9OQ+jaJZdcQnJKct/W8Xp0pNnLtPz8mG2JvWvXLoKBIEn5E2IyXiyNsZlw9/REZQe/wsJCDGYTlrHpET+3FtKsRgwKMZkdfPW111CMRlLmxmYNRaQsy0nGoCh88MEHWocywCuvvILRZsU+a3JUx0m3mvjG3Cysqp+fPvE41dX6/Ts6HDU1NZgcyWfducm0mVgwJon3N2ygu7tbo+j0qbOzs+/fvaVA4vz6l//o2XCm534BpEcpDnGRMJlMXHnFlXgrGwh59Ld4pd0bpKLTy6WXXRazMbdt24Y5LQVTpiNmY8ZK7yxwNGbNi4qLMGWlJcxdBqOikGIxRf22dGVlJTs//pikWXkD7iLEA4fFSEGmlT27d2kdSp/CwkKOHj2KvWBK1GbL+8u0mfj63CzweXni8cdjcocl2qqqqzE47IM+d8OkVHo8Ht5///0YR6VvHR0dff+WxHxo9N7Vqddw3tE+AB5yOp2D//YIMUTXX389aiiEW4ez5oebw7+4l8UoMW9ra8PlcmGdkhMXdb7DlWYNL55ta2uL6Hm9Xi+VFZVYxui7u89wqSoYonyh8c4772AwGUmZMyWq40RLuI1eC83N2pfDqarK3//+d0x2G8kzR98icajGJpn5akEmPZ0dPP6Tn8R1jbHX66W1peWs+vJeuSkW5mTa+Odbbw1IRi92/f+fS2I+NP3r8vVsOO8AucD/ATqdTmeD0+ms6P8RpfhEAsrNzWXmzJl4Sqp019P8QFMPuRMnkJsbm/Z7R44cAYiLTZdGIsUc/hMT6TfUqqoqQqGQ7ttuDpcKUb1A6+zsZNv27dimjsdgjc2OtpGWnxae5T9x4oTGkYTbdZ48eZLkBdNRTLHp4NRrYoqFR+Zk0tzUyM/+67/i5jb9mfoWfp5nrcPtU9Lo6enhpZdeilVYutfc8smFaXlZmXaBxJF4KYcaTmL+FPAF4GHgO8B/nPEhxJBdf/31+Du68dXFpjXcULR5g5xq97Js+eUxG/PIkSMYbdaELGOBTxLz/vWQkdDbucSUNvgsWzxSVRVfMITZHL0+9h999BEBvx+7M7q10NE0MdmMxWjQfLv2QCDAiy+9hDk9haR8bdqc5qdZ+bwzg7KyMp566indTXQMRW+d/Pl+l8cnm7l6QjIfffQRxcXFsQpNt0KhEF2d4XU7NqNCTW1N3MwGaynS70PRMuTE3OVy/dXlcv0VWAscAY4Db/Z7XIghW7p0KckpKbhPlGsdSp9DTW5UYNmyZTEb89jx45izMxKyjAXAYgh/XZGezauurkYxKBhTE6eyricYTsyjtehYVVU+/OgjFLOJjj3HI9b3PdYMisLYJFPM2kqeywcffEBDfT0pi5woBu1+fwuykrh9ahq7du1i3bp1msUxUtXV1aAomFLPf5F9U56DNKuJZ/70p4t+c53+ZVx5qRZUlYTq0hMt8VIKNeSVKk6n00J4AehXACOgAAGn0/ki8IjL5bq4f1PEsFgsFlbecAOvv/46gU43Jh0kWPubPORNmsSECbHpjuJ2u2ltaSF18syYjNdrW20XR1sGLoKZmGwZdMdFgF313eyuH/wW4F356eSmDF4S8eThhnBtBpGvm66rq8OUmpwwCz8hvPAYICsrKyrnLysro662NiEuZsYlGamq0W6NSnd3N6++9hrWnCxd7D1w7cQUarv9vPrqq+Tl5cVsjUwkVFRUYE5LvuCmTDaTgc9MT+OPx6p44YUXePjhh2MUof70rynPS7VQ3ObF5XJRUFCgYVT6Fy9rMYazhPwXwM3A7cAOwsn55cBvgMeBf494dCKhrVy5kjfffJPuE+WkLZ6taSxNPQHKO7zcf+uVMRuzqqoKAFN6fPSRHoneG+sRT8zr6zEk2y584Ci0e4Phi4szROsiptsf3nRrzJjoJHrbt29HMRgYc/OyQevLe/u+D8Y+LRf7tNisuxiKTKuJQzWtqKqqyd2mdevW4e7uZsw1C3Rxt0tRFO6dkUFdT4A/Pv0006dPJz09XeuwhuTkqVMY01OHdOyczCSunZjCBx98wNy5c1myZEmUo9OniopPlvUlmwzkOazs27uXVatWaRiV/sVqV+XRGk5ifh/waZfLtbnfY287nU438BKSmIthysjIYPny5Xy8exepC6ZjsESvtvZC9jeG6/OWL18eszH7EvMYb/By5fgUrs0d2hshwNLsZJZmD7+W+5vzx+EJhHjs4xpMpsi1kVNVlYb6eoyTx0XsnHrgCYQwGY1Mnhz5+m9VVdmzdy+W8Vlxu+izP18whM1q1SQprq6u5t133yVpei7mYawNCYVUitq8PHm4gRtyHczODF9YVnX5WHeyre+4h2Zl4bCEF5L2v9BzmI08NPuTuylrS9uo7g7fqJ6dYeOGSQ4+NzODnx9o5I9//CPf+c53dHHRcD4dHR20tbaSOtU55NfcOiWNkg4ff3z6aXJzc2N2h1NPKioqUExG1ED4Ltu8TCtvlZXR3NwctTtuiaAxSpvdRdpw3i0NwGANU5uBxJ3yE1F1yy23sH37dtyuClLmTdMkBlVV2dfoYZbTGbXZysFUV1djMBkxpiTFbMxY8wTDs8BJSZH7Gv1+f3gr+Sj34E6zGvnm/OEl/6O5iPnlwQayx0/BYol84lxbW0tjQwOOJXMifm4tdAdCpKTE/m1HVVWef/55MBlJXTgj5uNfSLbdzG1TUll38CAffvgh1113ndYhndfJkycBMGcN/QLHZFD4/KxMfnWokZ/910/54Y9+jMORmIvnz+XkqVMY01IINIdLM+ZlJfFWWQd79+7lpptu0jg6/YrF5m2RMJzEfCPwX06nc7XL5WoHcDqd6cATwKYoxCYuAlOnTmX+ggUcO1EY1e2sz6eyy0+928dtV8aujAXOvdtdIvEEwsUsdnvk6po9Hg8AigY/K9HS7Q9R1eXn9tnRKek6cOAAANYoteVcd7KNJNPAcqWorVsAQMExJvbJ2L59+zh8+DCpl84a9uZMBoPCzHQrD84aOKOZm2I55wXg+S70zvW9XTEhhSPNHl568UWWL18e0YviSCspKQFFGXbb0yybiS/NzuR3R5r5xc9/zv/v+9+PygWtHnV1dVFdXY19Zl5fYp5tNzMxxcKHmzZx4403JvR7ynD5/f6+f7d3dOB2uyP6fhQNwyn8/DYwE6h2Op0HnU7nQaAKmAA8GoXYxEXirjvvJOjx4S6u0mT83fXdmE2mmJaxAFRUVmA4x6YaicIdCM+YR/IPYW+Hl1j3jY6mA41uQqrK0qWD13iPVmFhIWZHMqYEuDujqlDe6WOmc+jlD5Hg9Xp57q9/xZyeSvKs2G0mNFwGReH2qWl0dXezYcMGrcM5r6KiIszpqSOakJnisLJ6ZjrFJSX87ne/IxQKRSFC/SkqKgJVxTI2fcDjV+bYqaisDD8v+pzZiaWyslKjSIZuOL8NPUAB4QWgs05/fgL4wOVyxV/zVKEbTqeTWbNmUVJYRrJzEooxdgmXP6Syr8nD4iVLY3oV3dHRQWtLK6lTYptcxFpDTwCAnJyciJ0zEAifM5E6suxucJM3KZcpU6ZE/NyhUAiXy4UpJz3i5+51V346M9KHvhh3NCU/R5t7eOZ4MwsWLBj260fj9ddfp6W5mcwbl+j+Z29yani3zLfWr2flypW6nCEMhUIUl5Rgzhv5XZyFY+3c6Qvy+p49PPPMMzzyyCMJP1tcWFiIYjBgOqP8Z9E4O2+WdbJhwwacMb5o1bPexNykQEAN1+fr/fsznL8uh4B5LpfrTZfL9TOXy/Wky+V6X5JyEQmrVq0i4PbEfNb8SHMPPf4gV199dUzHLS8P928fzuKxeFTr9mMxmxk7NnIlFL0LSVU1MWbIarp9VHT6WHFVdH4Ga2tr6e7uxjIuOv3RY+1ocw9Wi4VZs2bFbMz6+nre+udbJE2dgDU7M2bjjsbNeQ663W62bt2qdSiDqqiowOvxYBk7up/LayamcuOkVD766CP+/ve/x+UmS8Nx4OBBLOMyzrpjaDUaWDIuid27dsVNLXUs9LZIdFiM2M1GyuJgl9ThJOYeILqrrcRFq6CgAKfTifvYqb6V5rGwq85NVmZmzPu/9t5OM2UMvTtKPKrr9jNhwoSItkvsS8yDiZGYf1DZic1qjdrFYd+W5wnws9bpC7K3sYfll18e1R1Sz/SPf/wDFYXURbHdc2A0JqVaGGc3c2D/fq1DGZTL5QLAHIELxpsnO1gxIYW3336btWvXjvp8etXU1ERNdTWWiYM3Kbg2NwUFNS43moqW3osURYFJyWZK4mDn2OGUsrwLbHA6ne8AZYQT9T4ul+v/RjAucZFRFIV77rmHH//4x7iLK0mePSXqYzZ7AhS1eVh1960R77N9IQ0NDaAotG4+wJl3Xk0ZjnP2dXeXVtFTOvjGKo7LZp9zBr55w66YJ7IhVaWyO8CShVMjel5jb6lTKP5nxhp7Ahxo6uG2226PWpeR1tZWAIz26PZ9j4WtNV0EQiq33nprzMYsLCxkz549pCycEXffw9npVnYUFuL1erFa9TWvVlhYiCk5CWME9iNQFIW78tPwBkO89tprqKrKpz/96QhEqS8HDx4EwHaORdzpVhOX5ySzZctm7rjjjoiWEMarhn4tEqc6LLxXWUV3dzfJyfpd3zWcbGQusBcYCywGVvT7iG07C5GQ5syZw+w5s+mO0az57vpuUIh5GQuEb40rRsNZSXkiaegJ4PYHmTkzsrOMiVRDurGqE6PRxM033xy1MZqbm1EMCgZbfHet8ARCbKtzc+mllzJx4sSYjKmqKi+++CKm5CRSYjBZEGmzM234AwFOnDihdSgDhEIhjh0/jjk7I2K/zwZF4b4ZGSzJtrN27VpeffXVhCtr2bNnD2ZHMsbzNA24YVIqRkXh1VdfjWFk+tU/Mc93WFBVKNb5rPlwZsyfB95yuVy67dDudDqtwJPAPYAX+KXL5fqZtlGJ4bjn0/fwwx/+kO6iClLmRHamtb+QqrK7oYd5c+fFtHd5r/b2dizZmWRed+mwXjfSHRizblxKwO2h8bWPhv3akSrrCG9+MmNGZPs9972Rx/mbbqcvyN4GN1dfe11Ud2n0+/0oRmPcX9C8W9FBTyDIXXfdFbMxjx49SklJCY6lc+KyC1CWLfwWf2ZnCq1VVVXR3dVF2vwpET1vb3KuAGvXriUUCnHPPffE/c8+gNvt5vjx49iceef9ehwWI9dMSOb9HTu47rrrmDMnMfYuGKmy8rK+f092WDAaFAoLC1m4cKFmMV3IcGbMfwHofaXafwPLgRuArwDfdzqd92kbkhiOWbNmMXfuXNzHThHyB6I2jqvNS6snwDXXXhu1Mc4nFOdJ5VCUtntJSU5m/PjxET1vb9lRvM+Gba3pIqiq3HLLLVEdx2g0xv1FTF23ny013VxzzbXk5+fHbNzXXnsNU3LSiC6G9cB4On8LBmO3bmcoDh06BIA1J/KTIgZF4TMzMliWk8zrr7+eMAtCDxw4QDAYxDbpwpuerZyUSlaSmWefeWZAH++LjdvtprHhk4WwVqOBKakWDp/++dOr4STmHwCfdzqd+uu7BDidzmTgEeDbLpdrn8vlegP4GdJjPe7cc8894b7mroqojbGrrpuU5GQuvXR4M9aRoqoqxP8kzjmpqkpJh485BQURr9/vW/QXx4s/vcEQ2+vcLFp0acQvXM5kMBhQ47jHs6qqvHayDZvNxmc+85mYjVtSUkJRURH22ZNRjPpuj3guxtMzq3pLzPft24c50xGR+vLBGBSFe6en9y0I/fOf/xz3fc537tyJyW7DfEb/8sFYjAY+nZ9GbV0d69evj35wOtXb/ax/zeisdCvlFRV93Vr0aDilLLnA3cD3nE5nM2cv/tR6x4UFhLvGbOv32DbgP5xOp9HlcunrL5M4pxkzZrBgwQKOHj+OfWYeBktkd3js8gc50uLhxpv+JaadHfozm0yoPfH9RnE+zZ4grZ5AVLrdmEwmFEWhu7gSb23T2c9HYfFspG2v7abbH+T222+P+lipqamoIZWQ14/Bqs3P+2jsrOumuM3Lww8/HNOt199//30MZhNJcTpbDtB9eq2Onnb/7OzspLi4mOS50b3zYVAUVuWnYTEobNy4EZ/Px1e+8pWYL/SPBLfbzcFDB7FOzx1yWc7sTBuXjEni9XXrWLRoUVT2SNC7wsJCgNN/98K/C7MybPyzvINDhw5x1VVXaRjduQ3nJ/Qp4AvAl4HHgB8BPwH+4/SH1sYDLS6Xq/8FQz1gAS5870foyj333EPQ68ddFPlZ870NboIhlWuuuSbi5x4qh8MBvuiV6mjtRGv41zAaibmiKFis1rjtyuILhviwupu5BQURXxg7mHHjwn/+gl3uqI8Vac2eAK+XdVBQMIfrrrsuZuN2dHTw8ccfY5s6IeITA7FU0x3+G5OXp/W82Sf27NmDqqpDKskYLeX0Lqi3THawdetWnnnmmbicOd+3bx/BQJCkKcPrsnL39HTsJoX/efK3eDyeC78gwRw9ehRLpgP63fGamGImzWpiz549GkZ2fsP5i/MP4AnCpSEmYCbwU8KXIV+KfGjDZie84LO/3s/11SdKXFB+fj7z58/nWOEJ7M68EW3ZPBhVVdlZ38O0/HwmTZoUkXOOhMPhQPUmbu3fgaYeJozPiVqZxoQJE6h2t5G1csmwXjfSxbORtL22m05fgFV33x2T8XoT80BXD+astJiMGQkhVeWl4lYUo5kvfzm2M527du0K7zDb1Ebzhl0kF+T3tajzt3TQsbew79j0FQsxJoXfYvrfkTHYrGRctTBmMQ+mptuHyWjUVdu8HTt2YHYkY4rh5mo35jkIqirvffQRFouFhx56KK4WhG7btg1TShLmMenDel2K2cjnZqTzh6N1PP/88zzyyCPRCVCHPB4PxSXF2GbmESiv7XvcoCgsyLKx49Ah3G63LnfFHc5fuh8CN57+6L30+i2wkPDCUK0NtgFS7+fxN1UkuOuuuwh6fPSURG430LJOH3XdPq67/vqInXMkUlNTCfR4E2JR0pnavUFOtntZfvkVUXvzm5yXR7CtO+6+f13+IBuqupg7d27Mdq4cP348RqMRf7N+ayoH82FVF8VtXh548MGI7hw7FDt37gy3M43DTiz9lbT7mJQ3qW9TLq21trZyvLAQ6+ScmCfG/5Ln4JqJKWzYsIE1a9bEdOzRaG1t5ejRo9imThjR92xmho3rclP58MMP2blzZxQi1KdDhw4RDASxTjh7gfHCMUkEAgEOHDigQWQXNpzf1vuB1S6Xa7vT6VQBXC7XNqfT+TDwBvDVaAQ4DNVAhtPptLhcLt/px3IIz5q3aBeWGCmn08ns2bMpPn4K+8y8iCzA2lHbjc1qZfny5RGIcOTS0tJQg0HUQBAlQncD9OJgkxsVWLZsWdTGyMvLI/jRR4Tc3qgtIIuGd8o78AZDPPjggzEb02KxkD9tGuUNuu10e5ayDi//LO9g6dIlMS85a2tro/DECVLm5ZO64OxWn+ZMB1k3Lh30tee7IxMKqRS1eXny8MD/DxOTLayalj7oa3bVd4f3WxjEXfnp5KYM3pv+ycMN3Ds9g4pOH6vvuGLQY7SwY8cOUFWSpkZ3wfNgFEXhU1PT8AVV3nzzTXJycjQtZxyq7du3o6oqSfkTRnyOWyY7ONnh46mn/sCECRN0VdoULR9//DHGJCuW7MyznpvisJBuM7F161auuEI/vx+9hpPpjAEG+8veDehhZclBwAdc3u+xK4F9LpcrcYt5E9wdd9xBwO2hp9+tqJFyB0IcbPZw+RVXYLNpm8z1LmIL9ZxZfRX/9jT0MHlyXlQ3gXE6nQD46uPnmrum28eOum5WrryR3NzYltPMnjULX2MbTe/upHnDrgEf7XsKz/k6d2nVWcf3fvhbotMbuycQ4vmiNjIzM/jSlx6J+czqsWPHQFWx5cb30qQ9DW4UReHyyy+/8MExoKoqmzdvxjImHVNadHa5vRBFUbh7ejoz0238+dlnKSkp0SSOoVJVlY96v2fn2VToQowGhc/PziRJUfn5z/9bd33tI83j8XDgwAGsedkohrP/fhgUhSVjkzhy5DCNjY2DnEFbw22X+F2n09n7VapOpzONcN35pohHNkwul8sN/BX4vdPpXOJ0Ou8A/h34jbaRidGYP38+4ydMoOdExajLFvbWu/EHQzFdRHYuGRkZAAQTLDGv6fZR1eXjmmui2x9+8uTJ2O12vHXNUR0nUlRVZd3Jdux2O3fHqLa8v95NRtQo7g0QCaqqsqa4lVZvgEe/+b802Ta7uLgYg8mEKSM1ouc1GBRmplv55vxxAz7ONVsOsDQ7+azjez/ONVsO8JWCsexu6GH+/Hl9f2u0Vl5eTlVVFbZRzPxGglFReGhWBmkWA7/65S9obW3VNJ7zKS0tpaa6mqRpo5/kSLMYeXh2Jm0tLfz2N78Jr6FIUDt37sTv95M05dx3ZpblJIMKmzdvjmFkQzOcxPwbwHzCs+ZJwD+BKmAS8L8iH9qI/Cuwh/CFwlPAD10u18vahiRGQ1EUbrn5ZnzN7fgb20Z8HlVV2VHvJn/q1JhuUHIuWVlZAAS7E2ul/K56N0ajMeqlQgaDgTlz5hBoaIvqOJFyrMVDcZuXT3/6HlJSYj9bOGvWLEwmE9YJY8i6cemAj3O1loRwacaZx/d+RKO15I66bg429XDvvZ+JSceawbiKXJjGpKHEYVu9XltruujwBrjjjk9pHUqfDz/8EMVoPG+yFCvJZiNfnJ1Bd2cnzz7zjNbhnNPmzZsxmIzYhtmN5Vwmp1q4d3o6xwsLee655+Jujc5Qbdq0CXNaynl7vmfaTDgzbHy4aZPu+vwPubjV5XJVA0ucTud1wOzTr3UBG1wuly76D52eNX/o9IdIEFdeeSUvvvgi3YVlWMaNbPbnVEd40ecjN9wQ4ehG5pPEvEfjSCInEFLZ19jDokWLYtJvOj8/n7179xLyByLWtScaAiGV1091MGHCeK7XaNGxxWJh5syZFNdGbiF1pFV3+Vh3sp358+dz2223aRKDqqrU1tZi1qAGOlK6/SE+qO5i4cKFzJ597ouuWPJ6vWzdtg1bXvaQe+k/fawJgwI35DqYnRkuPazq8rHuZFvfMQ/NysJhCS/Q7V+P7zAbeWh2Vt9xa0vbqO4OLz2bnWHjhkkOJiRbuGlSKusPHODAgQNccsklkfhSI8br9bJ9+3ask7IxWCK3/8CS7GQaegJ8sGkT48aN44477ojYufWgqqqKkpISUi91XrAM7srxyTxzvJk9e/ZEdU3UcA373czlcm1CB6Ur4uJhtVq57rrreOuf/yTo9mC0D78+fEddN0k2m+aLPntZrVYyMjPoaevUOpSIOd7iocsXjNmCqt52l4G2Tixj9XG7fjA767pp6vHznUc/p2l3jHnz5nH8+HGCHh9G27nLIEZi3ck2kkyGESdR983M4K+uVlJSHXz961/XbBOYnp4efF4f1hH8jdGLDZUdeAJB7rvvPq1D6bNr1y48PT1kzpindSgDXD0xhXcrOvjrX5+joKAAiyWyvxejsWfPHjweD5nTI79W55bJDpo9AV566SXGjh2rm/fFSPjggw9QDAbs+Rf+vs3JtDEmycw777wd34m5EFq49tpreeutt+gprSZl3rRhvdYdCHGoqYerrr1O80Wf/c2cMZP9x45oHUbE7KzvJiM9jfnz58dkvN7OAv7W6CTm7d5gXweNkSac98/MZENVF86ZM1m4cGHEYxyO3jpzX30LSZP109caYN3Jdhp7/Pyfbz8a0909z9Rbb2y0x+fWF2UdXrbUdHHttdfpqvPGhg0bMKelDOuO51cKxmAzDbxAy02x8M35gy/KXZqdzNLswdcknKuO32RQ+NKcLP5wtJHNmzezcuXKIccXbZs3b8aUah+0q8hoGRSFz87MpN3bxB9+/3vS09N1c3dlNDweD1u2bME2OQfDECYfDIrCivF21hWXUFJSwvTp02MQ5YXFbxGduKiMHz+eOXPm0FNSPey6uAONbvwhlWuvje6CxOGaPn06/i53QiwA7fAFKWz1sOKqqzEaY9P7ecyYMVisVgJtXTEZbyS21YZrfe/9zGc039Bk6tSpWK1WfFFYMHtXfjrfnD+u7+IFPkmiej96L15g4KLGBWOS2FnXze233xGVnWKHw+sN/y4qOun7PRz+kMqLxe1kZmTw2c9+Vutw+pw8eZKTJ0+SNHOS5r8Dg3Fm2BifbGHbtq1ah9KnsbGRY8eOYcsfWe/yoTAbFL44J5Msq4Gf//d/U1ZWFpVxYmn79u14PB7sM4e+eeDS7GRsJgPvvPNOFCMbnvj76yMuWtdffz3Hn3wSX20T1glD33BkZ72bvEm5TJ06NYrRDV/vDIW3pgl7BFbdD9W22i6OtvREtHZzflYSqgorVqyI2dehKAoTJkygpj06rb/SrMZBZ+eGOmvnCYT40d565s2bp4vZKJPJxJyCAo4UnUBVVV0kSW3eIGtK25k2LZ9Pf/rTWofTr4Qm/hbFvVveQb3bx3e/+W1d7Wa4YcMGDCbTqPpwR9uisTb+WVxCY2NjzDezGsz27dsBov49SzYb+ercLH57uImfPvEE//n//l/UdmuONlVV2fD++5gzUs+76PNMNpOB5dl2Nu/ayf3338+YMWdvSBRrMmMu4sZll11Gkj2JnlND72le0+2nstPH1ddcq4tEpL+pU6eSnpGOpzJ+Nn45l32N0e9dPpjciRMJdQy+AYvWttZ00e0Pcs8992gdSp9LFi4k0OUmqIPvmaqqvFjcSlAx8I1vPKqL3Sl7/0bEW7eKk+1eNlV3cvXVV7NgwQKtw+nT1dXFjo8/xjZ1fEQXMEbaorHhC5m9e/dqHEn4Z2/Lli1YsjMxpUT/AivDauKrBVmEvG6eeOJxXbePPJ/S0lIqKypGdGfmqgkpoMK7774bpeiGRxJzETfMZjNLlyzFW9mAGhhae6N9DW4MBkWXu3spisLiyxbjr20a8tcTCVeOTxlx2UH/2XII127eNyO8w+AVV1wZs6+hV35+PgG3R3fdbXoCIT6s6WbRJZfopm4R6Ktz91Rrv6nG9tpuXK0eVq/+HDk5+qh5712Dovr11T7tfDyBEC8UtTF2zJiY7ig7FFu2bCHg9w+rtEALWTYTSWYj9fX1WofCyZMnqauri+nuqNl2M1+ek0VnWys/+6//wuOJvza+GzduDN+ZmTL8uwwZNhMLx9jYtHEjXV3al0ZKYi7iyvLlywn5A0NKLFRV5UCTh7kFczVdUHY+y5YtIxQI4qnQ/g1hpPY19KAoaLKyv28H0AZ9zfJsq+3C7Q9ytw7KM/obM2YMeZMn4y2r0zSOZk+AN8s6mD9vHjfopIUp0NdjPuT1axzJ0K092UarN8DXv/EoSUl62IQ7TFVVPvjgAyxj06PS8z7S0ixGXcwWf/zxxygGA7a82F6s5qVaeMiZQUVlBb//3e8IhXTRBXtIuru72bFjB9YpORgsI7vzdsMkB16vl7feeivC0Q2fJOYirhQUFJDqcOApu3A5S3mnjxaPn8t1OFvey+l0MmbsGHpKq7UOZURUVWV/Uw+zZs3u680eS3l5eVisFnw62mgoEFLZWutm7ty5ulvXALDiyivxNbcTaNdmZkhVVV4rbUMxmnjky1/WVYlZUlISiqKgen1ahzIkR5t72F3v5lOf+pRmGzKdS1FRUXjmd3qu1qEMicOs0NraomkMoVCIjz/+GMv4rCH3e4+kOZlJ3Dk1jb379rFmzZqYjz9SO3bswD/KOzMTks1cMjaJd999h7a2tsgFNwKSmIu4YjAYWHzZZfhqm1EvcEV/oKkHk9HIZZddFqPohs9gMHDN1dfgrWsm0OXWOpxhq+zy0+D2c+WVsS9jATAajcxyzsJfr+0ban8HGt10eAPceuutWocyqMsvvxxFUXCfrNFk/ENNPRxv8XDPvfdqcjF3PgaDgeSUFEJxkJi7AyFeKW1nUm4uq1at0jqcs4R3rTRh01lrznMJqmAyaVsHX1JSQmtrKzYNd0e9akIKl+cks379erZu1U+nmvP58MMPMWc4Rn1n5ubJDgL+AG+88UaEIhsZScxF3Jk/fz4hfwB/Y9t5jzve6qVg7lxddSgYzFVXXYWiKPQU63dXxnPZWdeN2WRi8eLFmsUwb948/O1dBLu1r4tUVZXNNd1MnDAhZv3chysjI4MFCxbgLa254MVtpHmDIdad6mDy5DxuuummmI49VA6Hg6BH/6Usb55qp8MX5Ctf/aouFs725/V6+XjnTqyTs3W9K29/XQGVtLQ0TWPYtWtXuIwld/CuT7GgKAp3T0tnWpqVPz/7LDU12lzAD1V5eTllZWUkTZ846rtvY5PMLMu288H771NRURGhCIdPEnMRdwoKClAMBrw1Tec8prHHT6Pbr/mmLkMxZswYFl6ykK7jp2h6bxfNG3YNqKH3t3TQvGFX30f/vufu0qq+x1u3HBxw3vY9hX3PdR09GfG4PYEQext7WH755X21uVqYNy+8m6C39tw/D7FS0eWnqsvHjTfdpKsSjTOtXLmSQI8n5msbttR00e4N8PnPfyFm/e6HKz0tDdWj7xnzolYPO+u6ue2228jPz9c6nLPs27cPr8ej6xaJ/amqSqcvqOlaJFVV2bV7V7iMZYR10pFiNCg84MzEqAZ58snf4vfr90J1y5YtKAZDxBbL3jrFgc2k8Owzz2hWZy+JuYg7ycnJTJs2DV/tuTdKOd4Snj2Nh8Qc4Ibrb4CQGhe30HvtaXDjC4Y03y1v0qRJONIcukjMt9V0YbNaddkFqL8FCxYwZuwY3EWVMRuz2x9iU1U3ixYt6lu0q0djx44lpLMuP/15gyFeKmknJ3scd999t9bhDGrbtm2YkpOismtlNDR5Arj9QXJztauHP3nyJC3NLdgmZ2sWQ3/pViOfnZFOeXkFL730ktbhDCoQCLBl61asuWMxWC+80+dQJJuN3DnFQXFJCZs2bYrIOYcrPu4xCXGGuQUFlLxRQsgfGPT5wlYvOdnZZGfr44/chfQmSh2GEFk3LhnwnDnTQdaNSwd9nX1aLvZpg7+ZpC0euKlNwB25Ug9VVdlW203+1KlMmzYtYucdCUVRWDB/ATv27EINqSgGbWaqu/0hDjZ5uOraa3VfPmUwGLjh+ht46aWXCLR3YUqL/h2PjVUdeIJBPvOZz0R9rNEYP348AbeHkM+vy97b/yzroMXj5//+769isUQmGYmkzs5ODh8+TNLsybq+a9SfqzV8F7L37psWdu/ejWJQNC1jOdPcrCRWjE/mnXfeYcmSJbq7oD548CDdXV1kLI7swufLxtnZ3eDmpRf/waJFi8jMjO0FpsyYi7g0Y8YMUFX8LWfv+hgIqZzs8DFfRxttXIjBYGDlDSvx1bfgb+vUOpwLcrV5qXf7uelf/kXrUIDwG2rQ4xv05yFW9jZ04w+FuP766zWLYTiuuuoqDAYD7pLodwTyBELsqHOzbNlyJk3Sd0/rCRPC5Rdada05n1MdXrbWdLFy5UpmzZqldTiD2rt3L6FQiKQ4WfQJ4GrzMCYrU7OJHFVV+XjnTiw5WRGb+Y2U26amkW4z8Zc//5lgUF/9/bdv347RZsU6IbK7dSqKwr3TM/D7fDz77DMx33BMZsxFXOrdtGWwBaAVnT58wRAFBQUxjmp0rr76al555RXcrgrSluo79s3VXaQ5HCxbtkzrUIDwgmAUBW91A5YxsV/ApaoqO+t7yJ86lSlTpsR8/JFIT09n0aJFHDh6hNSFM1CM0Zun2VnfjScQ4pZbbonaGJEyY8YMAFq3HsaUYhvwnCnDcdadqF7u0qpztj11XDb4a4YjEFJZU9JOZmYG991336jPFy07d+7EnGrHFAe9ywHavEGOt3pZeeM1ms3wl5eX09TYSNqyucN+7bbaLo629HBDrqNv07iqLh/rTrb1HfPQrKy+TeN21Xezuz6886/DbBywadza0jaqu8PllLMzbNwwyYHVaODOKQ6eO1HJBx98oJtF2z09Pezbvx9r/ngUQ+T/do1NMnFLXipvHDjI9u3bY9p5TGbMRVxKTU0lOycb3yCJeVGbF0WB2bNH/2YYSw6Hg8svvxzPyRpdb3BS7/ZT2Oph5Y03Yjbr41a/w+Fg+vTp+Kq1qTOv6PJT2+3jujiZLe919dVXE/R4z7uQerRCqsqWGjfOmTM1L3sairS0NKxWK6pPX7+DH1Z1Utft4wsPf1FXGwn1193dzbFjx7BMyo6bMpYtNZ2EVPgXDe/+7dy5ExQF2yT9lLH0t2BMEklGhVdeXkNnpz7u6O7bt4+A309SFFtLXj0xhSkOK3/963Mx7W0uM+Yibk3Ln8buwwcxpicPeLyk3UteXp6mnUJG6uabb2bLli24S6pIKdDf5jQAW2u6MBmNuivZuHTRIkrWrCHY7cGYbLvwCyJob324baRe7iAM1YIFC0hOSabnVE3UkoITrR5aPH4+p5Oyp6G46aabWP/WejKuWTTkOvPzrfcACIVUitq8PHm4Ydizm/6gSo07wJIlS1i0aNHIv7AoO3bsGKFQKCI/S08fa+LM5SITky2smpY+6PH9Z4LPdFd+OrkpZ5eIeAIhPq7rYcmSJYwbp01SrKoqO3fuxJqTicE2/DKWK8encG1u6oDHclMsfHP+4F/P0uxklmYnD/rcub63iqLw6Pyx/PeBBrZu3aqLO1979uzBZLdhHpsetTEMisL9M9L57wMNPPPMM/zbv/1bTC44ZcZcxK3c3FwCXW7U4CctjYIhlfIuP7Nmxddsea/Jkycze85selwVA74uvXD7Q+xu6OHyK67QvOfvmXo3kvJUxrYFYDCksr/Zw6JLL9X9os8zmUwmLl9+Od6qRkJRmiHeWecmNSVF1xt9nenSSy9FDal4qxovfHAMNHkCmMwWHnroIa1DOa9Dhw6BotBxoGjEbV/9zbFbJ/J+ZSc9gSC33357zMY8U3l5OQ0NDbrfiGliioUpDiubNn4Q85rrM/l8Pg4dPoQld2zUE+Vsu5lbJzvYv39/zDZckhlzEbcmTpwIQKhfz+Gabj/+YKivTjQe3X7b7fzsZz+j52Q19hn6Wii3s74bXzCk6W3fc5k4cSITJk6k9mAxnoo6AJIL8rFNHAuEE4OOvYV9x6evWIgxyQoMrA822KzDGtfV5qHbF9Rs99PRWrFiBe+//z6e8nrsMyLbLq7TF+Roi4ebb7lFd5vgnM+0adNIS0/HU1kfsV7cBoPCzHQrD84auNvphWY3U80G/nismc/eezcZGRkRiSVaDh85gmIxEYlc6SsFY7CZhj53eL6Z4ME09QTYXNPFihUrNO0Fv2vXrtNlLPrvILY8286LxXW4XC5NFx8fP34cn9dHcow62Fw9MYUjLR7++txzFBQURH3HYpkxF3Grt+dsqN/MS1lnOEmP58R8wYIFTM2fSvfRUzHfmfF8giGVLbXdzJ49W7cLHC9fvhzVH4jp3YaDTT3Yk2y63enzQqZNm0bO+PH0nIx8d5bDzT2EVJWrrroq4ueOJoPBwNIlS/DVNKEGtOtEEQipvH6qg5zsbF1eDPfX0dFBc1MTKQX5ZN24lKwbl/ZdFMMnbV97P3oviiFcBtT7uDkrNotG3zzVjsFk0rR9p6qq7Nq1C2v2yMpYYm3h2CQsRgMff/yxpnEcPXoUxWjAmhObNoYGReGzMzII+n089dQfor7xkCTmIm6NGzcOg8EwYKFkeaeP9DQHY8ZEtn1SLCmKwt2r7ibQ5T5nlwctHGzqoc0T4LbbbtM6lHNaujTc7z1pyvgRJwYZVy0c8njBkMrRFi+XLLpUNwthh0tRFK65+mp8Da0EOgav0R2pY80exo0do+nGLSN16aWXEgoE8dadeyOzaPu4rpsGt58HHnxQ93ccTp06BRCzxHo0TrR6ONzcw6c+dWfMe1T3V11dTV1dHdY8/c+WA1iNBnKTzZSVlWkax7HjxzCPSUeJ4e7BY5JM3DXVwbFjx3nnnXeiOpYk5iJuGY1G0tLToV+9W7U7wJSp+XHTEeBcLrnkEvLz8+k+chJVB71jVVXlw+ouJozPYYGO+8NPnDiRSXmT8JTXxWS8knYvbn+QJUuWXPhgHestw2n+YM+waoP9Te3nPKc3GKKo3cuiSy+Ly9/HWbNmYbFaNasz9wVV3q/qYpbTGRc7GPcma2adt0n0h1ReLQ3vnKr1JMPu3bsBsMVJYg4wIdlERXm5ZtvVu91uKsorsGTHvqxrWU4yc7OSWPPSS1RUVERtHEnMRVwb229mPKiq1Hf7ycvL0zCiyFAUhfvvv59Adw/druj9ARiqknYvVV0+brn1NgxR6BkbSVdcfgW+xjYCXe6oj3W0uQeL2Ry3ZSy9MjMzsdvthDw+IrWuq7TdSyCkcskll0TmhDFmNptZMH8+vupGTRa77ajrosMb4J57742LC5v6+npMdpsud0vt74PKDpp6/Hzh4S9qfpdrz949WMamD7h7p3cTk814fT4aGho0Gb+kpARVVbGMi31irigK981Ix25S+J8nn8Tn8134RSOg73dYIS6gf8lKvTtASFUTIjEHKCgoYN68ebiPnopax4yh+rC6C0dqalwscOxtWegpi+6suaqqHGv1MnfeXKzW+HljPZfPfe5zEAzhWOQccgmQ+TybORW3eTEZjcycGdntsmNp4cKFBNweAm2x3QXUH1LZWNVNQcGcuNmPoaGxAUOM25QOV4Pbz8aqLpYvX868efM0jaW5uZnysnKsMVrAGCmpp1t5ut3Rn/gYzMmTJwEwZ2nTFSzFbOT+GelUVVezZs2aqIwhibmIa/27FNS6w8lroiTmAPfffz9Br4+uoyc1i6HO7ed4i4cbb7oJi0X/C5TGjRvH9OnT8ZTVRnWcWneAFk+ARYsujeo4sbJ48WIMBgOeqsjMhJV0+Jg+fXpcX7T0lm15a2JbznKoyU2nL8Add3wqpuOORmNjo64Tc1VVeaW0DYvVygMPPKB1OOzfvx9At5sKnUvw9M0jYwzru/s7efIkZkeypndmZmfYWDE+mXfeeYcjR45E/PySmIu45nB8Us/Y2BNAURSys+OnXu9CpkyZwhVXXIH7RDnB7h5NYthc3YXZZOKGG27QZPyRuOKKK/C3duJvjd4udYUtHoC4qP8diuTkZGbMmIEvAruA9gRCVHX5mFNQEIHItJOZmUnupEkxrzPfXusmJzubgjj6/nV1dWGw6vfCfV+jm+I2L5+5737S09O1DocDBw5gTk3GlBZfG+EFT5d1aZWYl54sxaiDdQy3T00j227mqT/8ga6uyN5Rk8RcxLXU1E92PGvqCZCVmaF53WCk3XvvvRgVAx0HimI+ticQYl9jeEOh/hdBerds2TIUgyGqs+YnWj1Mys3VtKtDpF1yySX4WzoIuj2jOk95Z7hW3el0Rigy7SxZvBhfQ+uARa/RVN3l41SHlxtWrtT9eo5eoVAIT49Ht/Xl3f4Qr5/qYNq0fF3sWOzz+Th27BjmCdHthx0NnkB40acWd8I6OztpbWnVxQJji9HA55wZtLe387e//S2i546P33ohzqH/7pONniA54yOzGYiejB07lltvvRXPqVp89S0xHXtfoxtfMKSLN7PhSEtLY87s2Xgr6qOycM8bDHGy08eCBJkt79W7iNVbO7oWgac6vCiKwvTp0yMRlqYWL14MgDdCJT4XsrvejclojKve716vF1VVUcz6bOm4vqwddyDEI498WRcXO4WFhfj9fqz91nLEi7JOHynJyZq0JC4vLwf00/lnUoqF63NT2LZtG8ePH4/YebX/CRViFJKTP9nprckTSKgylv4+9alPkZGZSceewphuOrSjzk3epElMmzYtZmNGyrJly/B3dBOIQjnLyXYfwZCq+QKySMvLyyM1NRVv7ejKWco6fEzKzSUpKSlCkWknLy+PsePGRX0xMYTLBPY3ebhk0SJSUuKnxKHv4leH3WNOtnvZWdfNzTffopv1R0ePHkUxGLBmx9/dtlOdfmY6nZp0CupLzDNSL3Bk7Kyc5CArycyf//wsgUAgIueUxFzEtf6JeTCkRn2rXK3YbDYefOAB/K2duGPUPrGqy0d1l4/rrr8+Ltq1nWnx4sUoioKnsj7i5y5p92I0GOJ6h9nBGAwG5s2bh7+uZcR3GlRVpao7wLQEmC2HcIu0q6+6Cm9dc9TXebhaPXT6AnHR/WhwsW8reT6BkMrLpe1kZWZy9913ax1OnyNHjmAem45i0qZOe6Q6fUEa3X7NStQqKyvDLTl1tEuqxaiwKt9BTU0t//znPyNyTknMRVzrn5gDCVXve6YlS5Ywb948ug6VjLoGeCh6b6lffvnlUR8rGhwOR3gxYxQW7p3s8DF16lRsNv12oRip+fPnE+zxEmjpGNHrW71Buv1BpkyZEtnANLRixQoA3KU1UR1nf2MP9qSkuFtQ3Hfhrq+8nI1VndR1+3j4i1/Uze9qR0cHFRUVBLp6hrWZl7u0qu/x1i0HNYgcDjeHL0y1ulNYXlGOIS35wgfGWEFmEnOzknjjjdfp6BjZ383+JDEXcc1utw/4PJETc0VRePjhhzGg0LGnMKpjBUIq+5p6WHTppXF1S/1Ml156Kb6WDoLdkbuQCYZUKrt8zEyAhY2D6W0R6Blhd5bq7nDb0kRKzMeOHcvsObPxnKyO2mZD/pDK0RYPS5YujbsF7L3xqkFtdoMcTL3bz/uVnSxbtkxXm1y5XC4ADBZ91uOfz76GHiZOmMDkyZNjPnYoFKK6ugZz+sjej9q9QZ483MCThxv6OmpB+M5w7+NPHm6gw/fJTtu76rv7Hv9r4cB1N2tL2/qe+6Cyg9smO/B6vKxfv35kX2A/kpiLuHbmG1j/vuaJKDs7m1V33YWnon7ATEukudo8dPuCfTOF8WrRokUAEevNDVDfEyAQUhMq8ewvLS2NKVOn4hvhz1eDO1xnOXHixEiGpblrr7mWQKcbX0NrVM5/otWDJxBi6dKlUTl/NJlMJoxGI2qEamxHS1VVXi1tw2K18eCDD2odzgDFxcUoBgNjbl5G1o1Lh7yZl31abt/jGVctjHncLZ4AJzu8XHHllZqUNjY3NxPw+3XbXjIn2cxl4+xseO89mptHt3g+/i7ZhOjnzD8Q/bu0JKrbbruNzVs207L3BNacLBRj5K+vD5y+pd47exqvJkyYwLjsbNqrGkh2RmbhV1VXeBvmRE3MARYuWEDZG28Q8gWGPbPX6AmQmpJy1t2seLd48WJsNhs9JVVRWbR3qKmH5GR7XPUu789itaAGghc+MAYON3sobvPyhS98QRc9y/s74XJhznKgRKgP+LbaLo62DFz7MDHZwqpp6YMev6u+m9313YM+d1d+Orkpg9dvP3k4fKF+xRVXjDzYUaitDbe+NTpGVsqSZjXyzflnb+aUm2IZ9HGApdnJLM0efLzBvr//MtnB/qZ61q9fz+c///kRxQkyYy4STKIlA4MxmUx8/qHP4+/opruwLOLn772lvnjJEkym+L52VxSFxZddhq+uhZA/MrN5td1+zCYT48ePj8j59KigoABUFV/j8GeHG3sC5CTg98ZqtXL55ZfjraiP2M9SL1VVKWr3MX/+grj9nUtKsqNG+PsyEv6QyptlHeROnMh1112ndTgDBAIByk6dwjwmXetQhiUQUunwBVm4cCFjx2rT4rE3MTel6vc9Pstm4pIxSWzZvBm32z3i88TnXwAhBmE2m3XRozYWFixYwKWXXcqBg4dImjZxwC3P0SpsCd9SX7ZsWcTOqaVFixbxz3/+E291I0lTRp8w1rkDjB8/XrOd72JhxowZGE1GfHXNA261D0WTJ8iCBEzMITxbuGnTJrxVDSRNjdyeCbXuAB3eQFy330xJTqbbq83uxP1tremiucfP9x58UHe/ozU1NQQCAVKyIteH+8rxKVybO/T2geebBT6X/Y1ugircfPPNww0vYpqamlCMRgwRfK+LhhUTUtjb0MDWrVu56aabRnSOiyOLEReF1BT9rdaOps/e/1nUUIiuI6URPe+BRjepKclxe0v9TE6nk5TUVDwVkWmbWO8JkjtpUkTOpVcWi4Vp+dPwN7YN63XeYIh2b4CcnJzoBKYxp9NJekY6Paciu6NsUVt4MdrcuXMjet5YSk1NRfX5NY2hJxDig6ouFiyYr8uLnFOnTgH62SBnKFRVZUtNNxMnTND057OxsRFTSpLuW/dOTrWQl2rhvXffHfFCcUnMRcKwJ+tzUUi0jB8/nmuuvpqe4ioCXZGZqfIGQxxt9bJ02fK4vaV+JoPBEC5nqWlCDY6uBtYbDNHS40+4hY2DcTqd+Js7hvU9a+wJlzIkapmPwWDgisuvwFfbTCiCSWhJu5dxY8dqsptipGRmZhJyey98YBR9VN2J2x/k3ns/o2kc51JRUYHBZMSYGj+TSEeaPVR1+bj1tts0TYobGxtR7PqeLe91eU4ydfX1fRsiDZck5iJhXAz15WdatWoVRoMhYrPmR5s9+IMhli9fHpHz6cWSJUsI+QN4a0a3Wr7mdCtAvewgGE3Tp09HDYXwNw+9L2+LJ5zEa1WHGguLFy9GDYXwRqg/fkhVOdXhZ9bs2RE5n1aysrIIuD0x3Zm4v25/iM013SxevJipU6dqEsOF1NXVYUy1oxj0PevbK6SqvFPRSU52tuYdupqamzEmx8dOwgWZNhRg//79I3q9JOYiYVyMiXlWVhbXXnstnpM1EenVfajJTUZ6mmY7u0VLQUEBtqSkUe8CejEl5jNnzgTA19Q25Ne0ecOJeaLuwAvhC5a09HR6IlQa1eAO0O0PMmvWrIicTytjxowBVSUYoVnzp481Davn9NaaLjyBEJ/+9KcjMn401NTWYtTx4sUz7W/sobbbxz333qtpvX4gEKCzowOjXR+bRF1IqsVInsPKvr17R/R6ScxFwtDLzm6xdttttwHQXXhqVOfxBkMUtvpYsnRZwi2iNZlMLLrkEnzVjaOa0avp9pNks8V1ycFQpaWlkZGZib+pfcivafMFMJtMOBzxU0M7XAaDgSWLF+OvHX1pFEBZZ7j9Zu+FULyaMCG8GDbQ3hXzsb3BEFtru7nkkkuYpNP1H6qq0tTUiDElPhLzQEjl3YpO8iblat5bv7U13B3KGCelLAAFGVZOlZWNaCfQxCgiFYKLNzEfO3Ysly9fzse7dpIybzoG68h2DTze4sEfCrFkyZIIR6gPixcvZseOHfgaWrHmjGxGt6Y7QF7eZN0vQIqUGdOnc6Dw6JCPb/cGSU9PS/jvzyWXXML7779P09sf9/2+JRfk93Ww8bd00LH3k91501csPOe5yjp9JNuT4r4uPzc3F4BAWxcMs5PPYL5SMAabaeAEwbl6Tu+qc9PtD3L77bePetxo8Xg8BANBDLbB+4Trzc66bpp6/Hznvvs1n6hpagrvQhwvpSwAuSnhvws1NTXDnqhIrGkxcVG7WBNzCM+ahwJB3CWVIz7HoaYe0hyp///27jxOsruu9/+rqrur932bXmZP+CaTkH0hkECC7BIIIApGRAVxg6tyuV53uQjX5V5xQREVubhduf5cLuAFUZQoIAYS2RO+kkC2SWYms/X0Vts55/fH91R3dXVtXVVdfarq/Xw85jEzdWr59ulTVZ/zOZ/v59t2ZSw5V1xxBV1dXaSO17bUvB8EPLGW4cAeLEe9V44ePUpmeQ0/la7q/ktpn4mJ9i1jyTl27Bg9iQR+qv4JoA8vZzh69KKWP5kZGhpiZHSU7Pnlpr6uHwR88sQqR48ejXQ50MqKu5IQT9SWOGmmlOfzsUdXuMQYrrrqqr0eDk8+6eZzdA21TmA+0+9+zydOnNjxY5Uxl7bR29s6l7ka7eDBgxy77DL+42sPMnjpIWI7zHBk/YD7z6d4xjNv2vPsyG7p7+/HGMPXH38Urt35yce5lEcy63dEfXlObnXTzNlleucqB9wXMgGHx8d3eVR7L5FIcOzYMe576AEmn7f9Mn9uafVKMn7AibU0Nx49uhvDbLojhw/zxS9/mTOrd2/b1j0+wuj1xSe4rj34GOsPHgcgc25npTBfP5/iybUMr6yxZ3Sz5BacifVGP+z65+MrLKezvOrVr47ECeNGYN5CGfOJvi66YrGNhZF2oj2/gaUjJRKtcYlwt3zri15Edi1ZU7/u/zifJJX1ue6663ZhZNFx9dVXkzm/jLe68/aSuYmfBzsoY577WTNnq6uTXE57jI6O7uaQIuPYpZeSOb+Ct177ZMdTaxmCgMjWRe/U4cOHCTyPGts31+STT6wwPDS053XQlXjhfISdJk2abTnt8Y/HV7ju2msjM+/hxIkTdA/2E+uK9r7LF4/FGEx0bVwp2Ynon7qJVKnTA/Mrr7ySickJVh84vuMVLr98Jklfb2/bLCpUSm7RkdSJswwc3Vkv8uMrGWKxzVraTjAyMsL4xDjrVQTmnh+wnvXaeuJnvkvD9obpk2drXlH25Jrr+94uffFzbQpHrnkKienqr5wMHF1k4Kh7X539xL1Vt6JcSnl89WySF7/4dnp6ol0i4ucmnUcgA13ORx++QNaHV3/nd+71UDbcc889eIHHmb+/u+Jcjtwq2PlXYfxkGnqav9+zflDTcdk6px8iFUT9g3m3xeNxbrv1NlJPnCa7slb144Ig4L5zKa648sq234eLi4sMDg2RPrHzfuZPrGWYmZ7puLkMhw8dxquibng16wKP4eHqlwdvZYcPH6anp2fHq6PmO7meIRaLtc1KqbnAfCe97+vxuVOrBAHceuutTXm9etS6CmQznVjN8JmTqzznuc+NzGRk3/dJp9PEuveuXWOtvCCoaaE+BebSNjo9Yw7wrGc9C2Ix1r/xeNWPOb6aYSmV5ZprrtnFkUVDPB7n8ssuI3Py3I6/KE+ueyx0ULY859ChQ2SWVgmy5VsDrmY6KzDv6upy+6aOIPT0usfkxETbfHZNTEwwPDJC5kz1LTZrFQQBnz21jnnKUyITRJazUasd4QD9g99coq+3j5e97GV7PZQNTz75JEEQMHzFRUw+78aNbDlszuXI/clly8FdhcndvhedcPwgIOMpYy4drl2WkK/H1NQUlxhD6qETVQeeXz2bJBZzpTCd4PLLLye7uo63XP1VBS8IOL2eaZuSg504ePAgBAGZc+Wz5svhYi+dUmMObrGh7LkLNffGP53MMjs72+BR7Z1YLMaRI0fInt39ziyPrWQ4tZbhmc961q6/ViNsfD/50QzM7zu7zv3nkrz8Fa+IVDnaI488ArjJw63kTDKLHwQ1vb8VmEvbUGDuPOMZzyCztEK2QiCV8x/nUxw8cLBjAqpcHX3qierLWc4lPbJ+0BKZuUbb7MxSPjO8HGbMo/SlvtuOHDmCn/Vc7+4anEn5zLRRYA5w+NAhMksrDVl8qZwvnlknHo+3zIT13MqZUSxp8fyAD37zArMzMzw/Yt1tHn74YQB6xob2eCQ788Sqmz9Sy5wkBebS8nKXCPdyyeAoueGGG4jH46w/VLlNU8YPeHg5zWWXX96EkUXD7OwsE5OTOwrMTyezG4/tNFNTU/QPDJA9Vz4wX8m4QKzTAnOoraZ6Peuzkm6vjDnkXWGp8WSlGkEQ8KUzSY5demnLlE7l2vkGmewej2S7fz2xysm1DN/1mtdELsH12GOP0TMy2HI15ifWXBcvBebS0aL2gbJXhoeHOXbsGOlHT1XMzjx0IUXWDzY6THSCWCzGFU99KtmTZwmqvKx8JgzMZ2a2rzrY7mKxGAcPHCBbob/0SsYnHo8zODjYpJHtvdnZWfr6+8mcOb/jxz657o6pdrsKk7vCkq2yxWYtTq1nObWW4foWWqU4977w0/UvStVI61mfv3t0mWOXXhrJeUaPPvYo8dHW+0x5YCnF4sJ8Tc0CFMlI21DGfNONN97IV77yFbLnV+gZL51R+uYFt6JjVPrVNsvll1/OXXfdRebsBRJTlUt4ziSzdHd1Md4Bi+cUs7CwwNe/+Q2CICi54MhKxmdocKBtF6gqJh6Pc9HRo9jjj+z4sbnAvF06suRMT0+T6O2tOCehHvefSwJEYlXKavX39xOLxQjS0cqY/9Njy6ymPb7zzjsjsZhQvmw2y8kTJ+k/dqju51pKebzrS6e23b4wmODlR8eKPubuk6t89uRq0W0vOzLG4lDxSaXJrM8DS2m+9ZZraxpr53yCStuL2ofKXrr22mshFiP5aPnFhh5ZSbNvdpahodaq36tXrs48/cTpqu5/LuUxMTHRUUFnvvn5ebxU2vUDLmE147dMWUEjGWPInLuw40xou16FicfjHDx4sOo5LrW4/1yK+bk5pqenK985IuLxOEPDw3UtSNVo51NZ7jq+wtOf/vSNsqwoOXnyJL7v091iGfOvnUviBwFXX311TY9XxlzaRqcGTcWMjY1x5PBhHjt+Gq64qOT9Hl3Jcvl1pbe3q9HRURYXF3ny5Fl4auXl0M+nfCYXppowsmian58HwLuwuqUlWb7VrM/wdGdMIM6Xu9qUOX2e3vnqA8VzKY/hocGN2uN2cujgQR6scIWlVmkv4MGlFM992lUNfd5mmJmZ4dGVs3s9jA3/8OgyfizOt3/7t+/1UIo6dcpluLuHBup+rtHeLt50xc5Ogm+cHeTG2Z2fFHzh9DqDgwNcfPHFO34sKGMu0rauueYa0qfP45XIci6lPJZS2UhmSprh2LFjZJ5cqqrV3fm0z+TkZBNGFU25cotsmRaTa9mg4668ABw9epRYLEZ6hwsNnU95TE6258neoUOH8DNZvJX1hj/3N8N5MblVfFvJ7MwMwUpyr4cBuPamd59c45ZbbonsVZtcYN411L/HI6neuVSWL51Jcuutt9VcXqvAXNqGMuZb5eovU48XL9d4ZMUF7EePVs4Yt6NLL70UP5ut2FEj6wecT2WYmmrPIKoaU1NTdHV1kb1QvN4SYC3rd2RgPjAwwOL+xR2vAHou7TPZpsfUwYMHgcotNmvx9aUUXfE4xpiGP/dum52dJbO6vuutJKvxz4+v4AUBt99++14PpaTTp08T64oTL3GVLoo+/cQqAfC85z2v5udQKYu0DdWYb3Xo0CEGBgdJnzjDwJH5bdsfXU5v1IN2oksuuQSA9KmzJKbHSt7vTDJLELTfJL2d6OrqYnpmhnPlAvOM11EdWfJdYi7h+F2fIPADYvHqPofOp7Jc2aaB+eLiIvF4nMyZJfoPNvZ98/WlNEeOHKG/v3WyqDkLCwsQBJz+u7uJ92zNpnaPjzB6ffHuWGsPPsb6g8e33BZ4tS1qBW5y4qefWOOGG26IdFeg8+fP093f1zLf7WnP5zMn1rj22mvrmv+gFKO0vFZ50zZbPB7nsmPHyJw4W7Rt4iMraRYX5tuyxrUao6Oj7JubI33yXNn7nW7T7hk7tbiwgH+heClL1nc98Ts1MD969Ch+Jkt2ufSJS771rE8y6zMxMbHLI9sbiUSCxf37G94yMesHHF9Jc0mLtnfN9bQOso3LmH/qiRXe9aVT3H92s0TmsZU07/rSqY0/F9Kbr3f3yVX+x+dPsp71uP32lzRsHLthaWmJWF/xzidR9M+Pr7Ca8Xjxi19c1/MoYy7Sxi6//HI+97nPuVrPrq3n4cfXPK5+amfWl+dceskl/MunP1U209nJiwvlm5+f595/v5fA94kVlI2tZV32rlMD882FhpboGa1cznM+5QKldp63cPTIEY7/66cbOgH0sZU0WT9o2fK7ubk54vE4fftnGLmm+lKcgaOLDBzdulBNdi3Jk391147H4AcBSymPS4yJ/Pyi8y0UmK9mPP7xsRWuufrqutsPK2Mu0sY2OkYU1L8upz2WU1kOHDiwB6OKjksuuQQvnSG7VHrxnDNJj95EoqNWtCxmfn6ewA/wikwATYaX1QcG6u+e0Irm5+fpSSSqXgF0OcxgtnNf/KNHj+Kl0kWPl1o9vOzmxVx0UWt2kurp6WF+fr6hrSRvnhviTVfMcOnE5kI2i0MJ3nTFzMafkcRm2Ux/V5xsAC/61m9t2Bh2y+rqKvFEa+SPP/7oMinP5zte9aq6n0uBubQNlbRst3//fnr7+kg/ubVc4/FV13O5U+vLc3LtrAr3T74zySzT09Mdf3zlWiZmi5SzpDxXKtWKdb+NEI/HWVxYIFvlMvTLGXci084ne7msdubMUsOe84nVDCPDQy1dAnT48GG8Ko+T3XDX46vMzExHcpXPQsnkOrGe6Afmy2mPTz6xys0338L+/fvrfj4F5iJtLLcyYbYgk/fEmgvMczWPnWp2dpah4eGyHTXOpX2mI9pOrJk2WiYWmQCaDEtZOjUwB/de8stMjs23nHEZ89HR9u37vri4SE9PD5nTDQzM17IsNiDw2UsHDx4ku5bck4WGHrqQ4qELKV74whdFvotZEAQk15MtEZjnOtzccccdDXm+aP9mRKRuhw8fdstj5/XrfnI9y0B/f1tn7KoRi8V4ysUXbztxyef6TbdvLXC1hoaGGBoeKhqYp3yXMe/UicTgOm5k15L4VSy5vpLxicfjbV3609XVxaHDh6su76kkCAJOrGdZXGztwDxXhrPT9pqN8C+Pr9Df18czn/nMpr/2Tnme5+YndNfWC7xZch1urrvu+oZ1uFFgLm2j00sNSjl06BCB7+PllSCcWssyNz+vfUZ44rK0gp/ZHlClPZ+1jALznLl9c0VrhtOeAvPcFQVvpXJN9UrGZ3hwMPJZy3pddPQo2XMXqlrEq5KltE8q67uWgy3s0KFDdHV1kT59vqmvez7l8YXTSW579rNb4spWJuOu6hZONI+az5xYZT3r8ZKXNK7DTbR/YpEq5FoBKsgsLjfBM3N+c8LR6ZQf6f61zXTo0CGAohOyltIuoGjnSXo7MTc3h19kNcd0OPmzr69v27ZOkVs9sdzqqDkrGY+R0fa/WnXkyBH8rEd2qboSn3JOrbtArdU/txKJBAcPHmx6xvyzJ1fxg4DnPve5TX3dWm0E5l3RDVOzfsBdj69y7NJLG9opKLo/scgOKTAvbt++fcRisY1Jexk/4Hwy0/F9uXMOHz4MFJ+kluue0c61wDuxb98+sqvr264upJQx3wjMq+lCsprxGR5p/2Nq4711tv468yfD9QRaPTAHt+pw5sxS01YA9YOAu0+tc9mxYy3T9tXL7ZsqF+zaC587tcZSKstLG1RbnqPAXKTNdXd3MzU9hRcufnI2mSVgM5DodOPj424CaJGMeSdM0tuJjXKNguCz07uygGsVOTwyUlXGfDXb3h1Zcvbt20d3d3fV3WrKObWeJZHoaemOLDmXXnopgeeTblLW/IGlFGfWM9x6221Neb1O4AcB/3R8hUMHD3L55Zc39LkVmEvbUMa8tLl9c26RIeB00gWbrZI52W2xWIwjhw+TPVskMA9LWRSYOxudWbYF5j49Pd10dUV7otZum9u3D6+KziwrGY/h4eEmjGhvxeNx5nfQRrKcJ9ez7JudbYvPeWMMsViM9MmzTXm9e06tMdDfx/XXX9+U12uEjRJVovn7vufUGk+uZXjpHXc0/JhUYC7SAaanp/HW3JLN51PukvDU1NReDilSDh8+THZpedtS2bmMeSdkN6uRO5krzJgnvYC+Di5jyZmfn8evkDH3g4C1DgnMAfYvLlZ1slLJ6aTPvrn5Boxo7w0ODnLw0CHSJ3Y/MPeDgPvPpbnyqqtJJFpjFU3YTLQFBHs8ku3Sns9HHlnmyOHDu3Kyo8Bc2kY7ZFJ2y+TkJIQt7ZbSHvF4XFngPEeOHCHwg23lLMtpn6HBwY7PBOf09/fTPzCAt7p1AmjS8+nvb9/Wf9VaWFggu57CT6VL3mc9696Hg4ODzRrWnpqZmSG7mqyrM4sfBJxZz7TVVb6nXn45mdNLRbtBNdKjKxmW09mWWFAo38Znrh+9wPyu4yucT2b5rte8Zlc6KykwF+kA+XWZGR/GRkfavlXbThw5cgTYPkltJeMx2gHdM3ZiemoKbzW55bZkNmjrntzVyrXyK9eFZD1cjKlTAvPp6WkIgm3HzE4spTy8IGirwPzYsWMEvk+mzKrDjfDVM+vEYjGuvPLKXX2dRssF5kHEAvMLaY9/PL7CddddxyWXXLIrr6FvZpEOUDhhamxM7f/yTUxMuAmgBYuhLGd8xtQqcYvp6Wn8ta1BVtoP6OvgiZ85ucA8vzVpofWwtWSnnMhsdKupor97KaeTLqs8PT3dkDFFgTGGeDxOapfLWe47l+IpF1/M0NDQrr5Oo3V3hyt+NqAHfiN95KElskGMV7/61bv2GgrMRTpAYR/u0bGxvRlIRMViMQ4fOoRXUMpyIRMwOjq2N4OKqNHRUfz17aUandwqMWdycpKeRKJsxjzZYRnz3FwWb6X2jPnZlJvr0U6BeV9fH0eOHiVzavcy5udTHo+tpLm6xcpYAHp6egCa1lKyGo+tpLn75BoveMELdrVtpwJzaRuqMS+tMDDXZMbtDhw4QOb8ypZa2OW0p1r8AmNjY3jJ1Laa4dwXaSeLx+PMz8+RXSrdhSRXY94prSUnJiaIxWLb5iXsxLmwk1S7rcB72bFjpHexzvz+s26fX3311bvy/LspHo/T09OzbUL+XgmCgL/+xhJDQ0Pc0eC+5YUUmIt0gMIgQIH5dgcPHiTw/Y2FmNJeQNrzFZgXyO0PP7k1a75x6bnDLcwvlO3MkgxLWTolMO/u7mZsfJxsHaUs51NZRkeG2+7k75JLLoEgIHP6/K48//3nUkxOTLC4uLgrz7/bEolEZALzr55N8o2lFK/89m/f9atdCsxFOkAsFmMg78NEgfl2uUuTuYWY1sKSA+2rrcbCMih/PbXl9lZqxbab5ubmyK6sl7wEn1uMqa+vr5nD2lNzc/vwl2vPmJ9Pe0xMtFe2HOCiiy4C2JWFhvwg4IELaZ56xRUtezW5t68vEoG5HwR85JFlZmdmuPXWW3f99ZTiEOkQI8PDrK26oLNTeijvxMbiOWHGPIyftK8K5MqiPAXmReUvwtQztv3YSYUZ804KzPfN7uO+++7nzN/fDcDgZUfoW3D14pmzF7hwz/2bdy7SLWo1GzDRhleuBgcHmV9Y4MwuBObHVzOsZTyOHTvW8Odulr6+Ptayu9tOshpfOL3O4ytpfuS1r2zKlUEF5iIdYnh4mBMnTgDKAhczODjI4NDQRsY8R4H5VrnA3F9TYF7M5iJM60UD87QXEIvFOmp/7du3D4KAwA+IxXeevV3NBhxs0/fhUy6+mE9+5l8JgqChme2vn3fvz8suu6xhz9ls/X19pB59eOOELl/3+Aij119a9HFrDz7G+oPHN/7vr6cgUVuBiB8E/N0jyywuzHPTTTfV9Bw7pcBcpEPk18V1SkeInZqenuaJ1a2dWbSvthodHXWT+daTxHo2v0LUlcWp1B4w6QX0JhItW15Qi/l5t2LnyLWGxPTWieg9EyNMPu/Gjf+f/cS92x6/km7flVIPHz7MXXfdhbeapHuocfMOHllOMz01uW3ifyvp7+9v2AJDSymPd33pFADPWRzh0gl3xeqxlTR/843zG/d77SWTjCRcD/W7T67yiceWObWW5U2ve0XT1v5QYC7SIfL7JndKD+Wdmpme5rh9csttrdb/d7d1d3czNDxMdi1F96gC80LDw8MkehN4K8VrqlOe3zETP3PyF14qDMwryfpuEnY7B+YA2bMXGhqYP76W5eCxww17vr0wMDBAz9DAlhO3qh53dJGBo5sTXk/+9V2wVlu7zqWUx+T4ODfccENNj6+FAnORDpEfDCgwL25qaorsvevE+ja7P2hfbTc+NsbJ9XUY3byaoMDcicViTE5OcbZEe0C3SmpnBebT09N0dXeRvVC6v3spq+Ek7HY9QT5w4ADxeJzM2Qv0HWjMyqYpz+fJtQzPPHSoIc+3VwYGBvDTmYY812hvF2+6Ymbb7YtDiaK3A+wf6mHdC7jjBS/YWIm0GdSVRaRD5AdOnZaxq9bExITrphFO0Ovp6W67Fm2NMDY2RlDQLlGB+aaZ6WmCghr8nKTn09/fWSd78XicmZnZmgLztUx7B+aJRIL5+XkyZ5Ya9pxPrGYIgP379zfsOffC8PAwXjJNEDSmnGWnPvn4KomeHm677bamvq4Cc2l5uTPZZtV/tar8wElBVHETExMABLll0/t0AlPM2NjYtj7mOqY2TU1NlVxQZ90LGGzTILOcxYXy/d1LuZB27fLG2ni14qNHj5I9u9ywAPTkmutk0qr9y3NGR0cJfJ8g3fzOLMmsz7+fXuempz+96SeFimSk5eVKDdRppLz8zG8nTTzbidxEqVwP6n6VsRQ1OjqqdollTE5OukxfkR7M617QkeVRs7OzZJfXdhx8LnVAYH7kyBG8ZApvtbY66EIn1zP0dHdvTERuVbnFzLxk8atPu+neJ9dIZX2+5Vu+pemvrcBcWp4y5dVR4FTZxpe/31nLpu/U8PCwy2SF9b+gjHm+3JUXr8iEs/VsZwbm09PTBJ6/bWGqSjolYw40bAXQE2tZ9u3b19S66N0wPe163ZeaSL1bgiDgMyfWOLB/ceN300yKaEQ6RKt/SDfDaMEiJn0KzIvKXZ0KMpsTs3Tit2ly0q1SWSwwT2W9jjzhqzXIWs34JBI9bb0g04EDB+jp6WnYCqAn1z0WWryMBTYX6/JqmJtQj0dXMjy2kubZ3/KcPbm6rMBcWt4VV1wBoEl6FTRjxbJW19fXRyIv89uJmc1q5FrX+Xm1nwrMN+Wyu4XZ4SCAjB905L4qd7JSzkrGZ7jNa/K7u7s5cvQo2QYE5inP5+x6puXry8F9zvT19ZGtYW5CPT5zwk36vPnmm5v6ujkKzKXlveY1r+Gnfuqn2nbWfqPoxKU6o6ObcxU6MbNZjWIZc5WybMoF5oV1+JmwRKoT99XGirE7LGVZzfoMtWkP83xPufhiMueWNyae1+rUepaA1p/4CW4u1OL+/WTPLVe+c4OkPJ9/f3KdG5/2tD1LzCgwl5Y3ODjIU5/61L0eRuSplKU6Y6NjG/9u58vn9cgF5vkZ804MNksZGBigq7trWxCa7uDAfHBwkO7ubrwSbSRLWc8GDA21f2B+5MgRAt8nU2cQeirsyJJbbbXVHTl8mOy5xnWsqeSLp9dJeX7TWyTmU2Au0iEUmFcnv85cGfPicvsoSCtjXkwsFmNkZHRbS8mU17mBeSwWY3RsdMcZ82SHdLHZmAB65nxdz3M26QLzXE1/qzt06BB+Jtu0OvPPP7nO1OQkxpimvF4xCsxFOoQC8+rkL/2tjHlxiUTCLTuvGvOSJsbHt5WypDs4MAcYG9u+TypZ94KOOEGenJxkcGiIzNn6MuZnUh6jw8Ntc4wdOXIEgPSZC7v+WisZD3s+xU1Pf/qethTWbDCRDqHAvDoKzKszPDzC2eXN1Qo1uXirsbExHjl7asttad/VD7dL0LRTE+PjPHL65I4es571GBwc3KURRUcsFuPQwYPcZ7/GmWWXHR687Ah9Cy7znTl7gQv33L9x/+HrLin6PGeTWaZmWr++PGdxcZFEb69rJXlkd8tzvnR6HT8IuOmmm3b1dSpRxlykQygwr07+JOJODaCqMToysqWURYtWbTU+Po5fUE+dzHZ2f/yxsTH8HXRl8fyAVNbviMAcXNlGkPWop5x6KRNsdMBpB/F43K2Menqp8p3r9JWzSWampzh48OCuv1Y5SnGIdAgF5tXJD8xVnlHayMgINGc+VkuamJjAS7nVP2Pd7r2XCjtudGpgPjk5iZfO4GeyxHsqhx/r4QJWnRKY79+/H4Cxp11G9+jWLmM9EyNMPu/Gjf9nS5zgXEh5Gx1w2sXFF13E1772NQLPI7ZL32Npz+fr51M8+7nP2vMkgzLmIh1CK6RWJz9oUovJ0oY7oIVdPXLBUX5NdZgw79gSqY1e5qvVLTK01mGBea7FYeb8Sk2PT3s+69n2C8w3O9bUtl+q8fWlFBk/4Oqrr96116iWvqlFOoQy5tVRYF4drRtQ3sTEBEDR0o1OLZGampoCwFutrpxlPZws2ymB+cLCAsRiZGsMzC+k3YlM4QrGre7w4cMAZM7uXjmLPZci0dPDpZdeumuvUS0F5iIdQhnz6uRnM1XKUlqnBEu12siYF+nb3anH1czMDADeijLmxfT29jI1NUV2qbbAfCXjAZvrDLSLqakpBgYHyZ7dvc4sD15Ic/HFF0ciGaNvapEOsdd1c60iP5sZhQ/pqOqUYKlWm6Us27PDnRqYj46O0tXdVXVgvp5xgXkn9DHPWZifx69xCfqVcH+1W2Aei8XYv3+R7Pnd6WW+nvV5fDWNuaR4p5tmU2Au0iEUmFcnP2hSYF6aAvPyBgYG6EkktnVm6YrHO7a1ZDweZ2Jicsc15p1UNjU3N0f2wlpNK12utmlgDnBg/wG8pZVdWQH0oQtpggAuUWAuIs2kUpbq5GfMOzWAqkYnZTFrEYvFGBsb25Yx79Rsec7M9DR+lTXmnVbKAi4w97PZHa+QCpsZ83acmL24uOg6+hQpDavXQ8spYrEYF110UcOfuxb6phYRyZOfJVfGvLT8YEkTi4ubmJjYFkh0emA+PT1ddS/z9axPT093R70P9+3bB0C2hnKWlYxHoqenLbv+zM3NAZC90PhyloeXMywuLERmvykwF+kQKmWpTn7gpIx5aQrMKxsfGyNIZbbc1tvhgfnU1BTZtSSB51W8b8oL6I9IsNQss7OzAHg1BKArGZ/h4fYs+5mfd6t+NjowD4KAR1YyXHTxxQ193nooMBfpEArMq5OfnVPAWVp+W0mVSRU3Ojq6rSShr7+zAs1CO2mZmPIC+no7a39NTU3R1d1VU8Z8NeszPNx+9eXgJlMnentr7lhTyumkx1rG4+jRow193nro01SkwyhALy8/yFRgXlp+jbn2U3EjIyN46QxBuOInQG9fZ676mbOTRYZSvk9fh62SGo/HmZmZqSkzvJ71GWzTibKxWIz5ubmGZ8yPr6QBOHToUEOftx4KzEU6xG7MZm93ygSXll+Pqf1UXK47hp9Kb9wWlTrWvTI9PQ1U18s87QX0duD+WlxYxL+w84z5utfek7IXFhZqbiVZymOrGbricfbv39/Q562HPk1FOoQC851TJri0/Csv2k/F5dr8+Xl15p0++XNycpJ4PF5dYO4HHblK6tzcHNnlVQLfr3znPEkvaOvAfH5+nuzKOn4m27DnPL6SYWFhPlITjBWYi4iUoICzOsqYF7cZmG9mzKMUAOyFrq4uxifGyVYVmNORgfn8/DyBH1S9EFPOWsZr657vuQmgtUyMLeXkusfi/gMNe75G0KepiEgJCjiro/1UXK6fdH7GvNMDc4B9s/vwq6gxT/udWfqz0YFkqfoANO35pD2/LXuY5ywsLAA72y/lpL2Ac8nMRivGqNCnqYhICcqYV0cTiosrljHv9FIWcL26veUqJn9m/Y4MzDd6du+gA8lqB6ySum/fPuLxeMM6s5xOZglAgbmISKtQwFkdZcyLywVJgWrMt5idncVLpvDTmbL3S3mdGZgPDg4yNDy8o5aJa5n2D8y7u7uZnpluWGeW0+uuVj23qFNU6NNUpENo8ufOKTCvjgLz4hKJBD09PVsCUJWy5K1uWSbA8nHlGZ0YmEN48rKDwHw96z7f8xf+akcL8wt4NXSsKeZcygXmud76UaFPU5EOoSBTdosC89IGBgfVlaVANTXUac8Fmp0amO+bnSWoYhGmnPWwV35/m/d9X1hYIHth5x1rijmX8ujp6Y5cXb4+TUVEpCa5Gnyd9JU2NDSIn86Qu2CljDnMzMxUrBVOhYFmpwbmMzMzZFbXqw5A18Ma83Zulwjuakvg+1WtHFvJ+ZTH5MRE5D6/FJiLiEhNuru7AU2SLWdocIggnQFcZK6Mea5WeIZsmZKEVIdnzKempiAI8NZSVd0/t7/aPWOeK4PyluuvMz+X8picmq77eRpNgbmIiNQkF5CrlKW0oaEhSHsbGXMF5s7+xUX8MsFVskNKM0qZnJwEqKqtJGxeYWj3/bU5P6H+OvPzaX9jP0eJPk1FOkTULtdJ61NgXtng4KAmfxYxPz9ftla4UzLApeQCxmyVJRspLyAej7f98TU2NkZPIrGjibHFeH7AhVQ2chM/Abr3egDVMsbcBPxrwc1ftNZeFW6fAH4PeD5wFvgFa+0fNXWQIiIdJBeQKzAvbXh4eEsf81z5T6ebm5vbWN2ye2R7J5FczXSnB+b+WnWBedIL6OvtbfsETCwWY2pqirNVXkkoZSntEQATExONGVgDtdInxGXAPcDtebflN0F9PzAEPAO4Hvg9Y8zXrbWFwbxIR8oFBO2eUZHmUca8sqGhIfyst/F/vf+cjUV0LqwWDcyTYfu/dp/MWEpfXx/9A/14VQbmKc+nv0Pq8WdnZjjxlS9z5u/vBmDwsiP0Lbha8czZC1y45/6N+47dclXR5zifcu/JKJaytFJgfgz4qrX2ROEGY8xRXMB+sbX2AeDLxpinAz/M9iy7SEfK9bft1AyUNJ4y5pUVtmJTxtyp1Ms8V2PeqYE5uGzu6WpLWbJBx3y2T09PE3j1tUtUYN4Yx4B/LrHtRuCJMCjP+RTwc7s+KpEW0+6XOqV5cgG5jqnSCldiVMbcGR4epq+/H2+5eElCssPbJQJMTU7x5GMPVXXfpOfTN9AZgXmuY834bdcS79kaxvZMjDD5vBsrPsf5tAvMVcpSn2PAmjHmK8AI8FHgJ6y1S8Ac8HjB/U8Ci80doohI51DGvDJlzIuLxWLMzs5wYqV4xjztBfR0d3f0/pqYmMCzX6vqvikfhvs74+pCLsvtrSaJjw1VuHdxy2mPRE9PJK8yROaIN8b0UTqQPgnsB/4deC0wCbwT+N/AtwIDQGGzzxSQMMbErLVai1xEpMGUMa+sMGPeyYFmoX2z+3j8vq8U3ZbyAnp7e5s8omiZnJzES6YIPI9YhbUCUn7AVIfsr1yW219LQo2B+UrGZ2RkOJKfXVH6hLgO+GSJbd8LjAPL1loPwBjzWuAeY8wBIAkUHpG9wLqCchGR3ZH7UgsCfcyWUlgjrcB80+zsLNnPfZbA94kVXHVJeT59fZ2RAS4lPzNcbIJsvnQHncjkAvNqJ8YWs5zxGBkbbdSQGioynxDW2k8BOzl1yU27XQCOA/sKtu8DnmjA0EREpAiVsFRWGJhrldRNbnn1wAWew1v3U9IL6BuMXplBM+V6bCsw32psbAyg6lVRi1nLBoyPjDRoRI3VEp+qxpjrjTHLxpj5vJuvBjzg68C/AQvGmEN5228ObxcRkV0QxcvAUdPX17dlPykw3zQ7OwtQdLGYpOdHsv63mTYy5muVe3a7KwydMVE2kUi4hbvWa8+Yp/zoTiyOTMa8gi/isuJ/aIx5C67G/PeA91lrTwOnjTEfA/7YGPNG4FrgTuC2vRqwiEi7U8a8slgsRqK3l1TSBREKzDflt0zsnd+6AmPKg9EOD8w3SjYqtEwMgoCM55NIJJoxrEgYGx/jdB0Z86QX3faSLfGpaq1NAy/CLSj0aeCvgb8H3pR3t+8GzgN3Az8PvN5a+5nmjlQk+pTllEbRsVSdvr7NEgPVmG8aGxujr6+vaC/zVIQDp2ZJJBIMj4xUDMwzfkBAdDPAu2F8bJwgWXtgns76kS39aZlPCGvtN4CXlNl+qtx2kU43ODhIb28vr3jFK/Z6KNImNPmzOv39AyydXwJ0lSFfLBZjbn6e40tL27Z1UmlGOdNTUzy2sn3/5Et77v0X1UBzN0xMTOB/88GaHusHAcmst7HoXtS0TGAuIvVJJBK8973v1aV0aRgF5tXJz/wqY77Vwvw8j35+e58GBebO1NQUj9pTZe+T8t37r5P21/j4ONm1JIEfEIvv7MpdMuuuMEQ1MNepu0gHUVAujaRSluoMKDAvaX5+nuzqOn4mu+V2P+isQLOU6elpsivrZU9+OzVjThDg11DOsh6uKlvYMSkqFJiLiEhNFJhXJz9jrpPjrebnXbO1YnXmnRRoljI5OUngefjJdMn7pMJAs5P212Yv850H5mtZt7+UMRcRkbakUpby8jO/Csy3mpubA8ArEpgrY761l3kpqTBj3kmTZTf3S+VWkoXWMgrMRUSkDSljXp38gEn7bKtcL/NskV7mCsyrC0BTHVjKsrFfVnYemK8rYy4iIu1IQWZ1FGCWlkgkGJ8Yx7ugwLyYagLztN95gfng4CD9A/0KzEVERHJyJSwK0MvrpICpFrMzs0UDLAXmYZvbvr6yAWiuxryTSlkApqdn8Fa2n9BVkvSi3cVGgbmIiNRFNeblKTAvb3Z2Fr9IRrjTAs1iYrEYU1OTZWvMk1n3/otql5HdMrdvH34NGfOoX2FQYC4iIjVRprw6UQ0AomJ6etr1pPa8LbdHNaPZbNNT0/jlAnPPJx6Pk0gkmjiqvTc7O+taSfo7SwykvYDurq7Iti5VYC4iIrKLenp69noIkVaq84gCc2d6erroFYWcpBfQ19vbcSfKMzMzBL6Pv176pKWYlOfT2xvdkxgF5iIiIruo0zKZOzU5OQmAt7Y1wNKVBmdqagovncFPZ4tuT2Z9+vs77yRmZmYGKN7Rp5yUF0T6pE+BuYiI1KTTMnS1iuol86jYWCxmVYF5MRtXFNaKZ81TXkB/X+fV4+cCc295Z3Xm6YjvLwXmIiIiu0gZ8/LGx8cB8Ne3ruKo/eZsXFEoUWee8nx6O3Ci7OTkJLFYbMeLDCUjvr8UmIuIiOwiZczL6+vro6+vD68gMNcqqU6lXuZJL+i4jizgjo+x8fEdB+brXhDZHuagwFxEROqkdonlafJnZaNjY/gFNeYqlXLGxsaIx+Mle5kn/c5tLTkzPV1DxjzarSUVmIuIiOwiZcwrm5yYwF9Pb/w/Hld4khOPx93qqCVKWTo1Yw6uY02wmqp8xzzrWV8ZcxERaV/KbJanwLyysbExguRmgNXdrTKWfNNT09uuKOSksl7HZsynpqbIrFbfyzwIAtYyngJzERGRTqVa6crGx8fJriUhjK90MrPV1NRU0UWGAiCV9SPd/m83TU5OQhBU3cs87Qd4gWrMRUSkjanGvDwF5pVNTEwQeP7G6p89Csy3mJqacquj+v6W29NeQEDnLsZUanGqUtaybv8NDQ3t2pjqpcBcRERkFykwryzXEjC3iE53tybM5puenoYgwFvbWk+dDgP1Tu35XmpxqlLWw8BcGXMREWlbqjEvT2UZleV6mQfpDAA96mG+xcaJS0EHkpTnrlZ1asa81OJUpSSzbn9FuSZfgbmIiMguUsa8slyA5WdcxlyLC22VC8yzBQFoLjCPcqC5m/r7+0n09ladMc/tryhfYVBgLiIiNVGmvDoKzCsbHR0lFosRKDAvqlTGPO250oxOzZjHYjHGx8e3rRpbSircX1E+kVFgLiIisosUmFfW3d3N8PAwhG3vtCjTVn19fQwMDm7LDCdbIAO82yYnJvDXqg3Mo1/6o8BcRERqkuvGoq4s5Skwr87I6OjGvxWYbzc5ObmtljoZZoA7dYEhcPMTgioz5skWuMKgwFxERGqSK2VRSUt5CsyrM5YXmGvC7HZTk5PbMsOtMJlxt01MTLhWklUkCFqhJl+BuYiIyC7SiUt1hoeHN/6tGvPtJiYmttVSp1ogA7zbJiYmCHwfP5mueN+UF9Dd1RXpEz8F5iIiIrtIgXl18ntLq5Rlu/HxcbxkisDbXGSoFWqmd9tGR58qOrMkPZ/+/mjvKwXmIiIisufy66QVmG+X6/Xu5WXNU15AT3d3R5dLTU9PA5BdWa9wT1f6098X3TIWUGAuIiIiEZAfmEe51GCvjI2NAWwpZwno7I4sADMzMwB41QTmnk9/xCfKKjAXERGRPZc/IU+B+Xa5jHlhnXlvb2fX4w8MDDA4OIi3vFrxvslswMCgAnMRERGRsvLrpDu5NKOUXMbcW99aS93J9eU58/PzZC+sVbzfuh8wMDBY8X57SYG5iIiI7DkF5uWNjIwQj8fx1goz5grMFxYW8JaqyJh7QeR7viswFxGRumiBIWmE/MBcpSzbxeNxRkZHt5eyKGPuAvNkCq9Cy8T1jK/AXERE2pMCcmmk/MA8Hld4UszE+DjemkpZCi0uLgKQXVopeR8/CEhmvS1tOaNIR76IiIjsOU3+rGxiYoJgfWtWuNO7skBeYH6+dGCe9AICUMZcREREpJL8AFM15sWNjY3hJwtrzBWYT0xM0NfXVzYwX8+6hZkUmIuISFvTypbSCCplqWxsbMzVUeet/qnA3H0GLSwskL1QJmOuwFxERESkOgrMK9tomZjaLGdRYO7Mz8/jL5deZCjpuTkxCsxFREREKsivK1cpS3Gjo6MABEkF5oXm5ubIrq7jZ7JFt+cy5vlzGaJIgbmIiIjsufySKGXMi8sF5vktExWYO3NzcwB4y8UXGsplzBWYi4i0mIMHD+71EEQ6mgLz4jYC87xSFrVLdGZnZwHIlgzMWyNjrn5EIiIF3vKWt7C6WnkVOXHUz1waTaUsxQ0PDwPgJTMbtyUSib0aTqTkAvNSGfN0mDGP+omMAnMRkQKTk5NMTk7u9TBEOpYy5sX19fWR6E1syZirlMXp7+9neGSkYsY86vtLR76IiIhEigLz0oaHh/HzJn/29PTs4WiiZXZmpmTGPOUF9CYSkT+2oj06ERGJrFzbsai3H5PWE/XgaS+NjIzip1TKUszs7CzBarLotmTWZyDi9eWgwFxERGp04MCBLX+LNIoC89JGhocJ0puBuerxN83MzJBZXSfIW4ApJ+kF9LdAEkFHvoiI1CQXPCmIkkbTMVXa8PAwQV7GXKUsm6anpyEI8Fa3LzS0nvUj35EFFJiLiIhIxCgwL21wcBA/rcC8mJmZGYCigXnSCxgcGmr2kHZMR76IiIhEigLz0oYKgksF5pump6cB8FaKZMy91pgPoyNfREREIiV/FVDZanBwcMv/FZhvmpiYIB6PlwjMfQXmIiIiIjuljHlpCsxL6+rqYmx8nGyxUhbVmIuIiIjsnALz0hSYlzczPY1f0DLRDyDjKTAXERERkQZSYF7e1NQUwVpqy20pLwDcyqlRp8BcREREIkUZ89IKA/Pu7u49Gkk0TU1NuV7m/mYv83TY11wZcxERaVu9vb1b/hZpFE3+LK0wMNe+2mpqasr1Ms/LmmddwrwlPqt0miUiIjW57bbbeOKJJ7jtttv2eigiHaMwMJetJicnAfDXkjC0NUOeSCT2Ykg7osBcRERqMjQ0xBve8Ia9Hoa0IZWylNYKweVeygXmbpGh8S3bWiFjriNfREREpIV0qa68pM3APLltmwJzERERkR1S3XR5Ay0wiXGv9Pf309ffj7e2PTBXVxYRERGRHVJgXl4rBJh7aWJiomhgroy5iIiIyA4pMC+vr1+BeTlTk5P4Bb3MoTVOaBSYi4iIiLSQvt7oB5h7aWJigmB9e2CujLmIiIjIDiljXp46s5Q3Pj5Odi25ZZEhaI39psBcREREIkEBeXVaIcDcSxMTEwD46+mN2xI9PS3RhjP6IxQRERGRDQrMyxsfd/3L8yeAtso+U2AuIiIi0kJ6enr2egiRtpkx36wz7+1VYC4iIiKyY0EQ7PUQIk2BeXnFMua9iehP/AQF5iIiIhIxqjUvr1srf5Y1PDxMPB7fEpj3qJRFRERERBpNGfPy4vE4o2OjW0pZWqGHOSgwFxERkYhRKUt5yphXNjExiZe3yFCiBXqYgwJzERERkZaijHllkwWLDLXC4kKgwFxEREQiRjXm5XV1de31ECJvfHx8SymL2iWKiIiI1EClLOUpY17Z+Pg4XjoDvjuWlDEXERERkYZTxryysbExAALPB5QxFxEREZFdoMmflY2Ojrp/eB6gwFxERESkJiplKU8Z88pygXkuY94qJzMKzEVERERaiALzyjYy5qFWqctXYC4iIiLSQhSYVzY8PLzl/62yzxSYi4iIiLSQeFzhWyXd3d309fdv/F8ZcxERERFpOAXm1RkaGtr4d6vUmLfGKEVERFrY85//fBYXF/d6GNImtABTdYaHhjj95JOAAnMREREJvfa1r93rIbSE7u5uMpmMMsIVtEq99F5rxYy5jnwRERGJhIGBAWD7xD3ZSicu1RkcHNz4t2rMRURERHZAJRrVUWBenfzAXBlzERERkRooQC9PgXl1cldgQBlzEREREdkFCsyr0692iSIiIiKymxSYVyc/Y94qE2b1mxURERFpIbnAXCU/5eVnzFVjLiIiIiINp4x5dfIDc2XMRURERKThlCmvjgJzEREREZEI6O3t3fi3AnMRERERkT3S19e38W8F5iIiIiLScCplqY4y5iIiIiIiEZBIJDb+rcBcRERERGSP5C8qpMBcREREZAcWFxcBtQOUxsjPmLfKMdUa3dZFRESk7b3uda/js5/9LKOjo3s9FGkD+VlyZcxFREREdmB2dpbbb799r4cReZr8WZ38/aTAXEREREQkAlqllKU1RikiIiIiUiMF5iIiIiIiEdAq5T8KzEVEREREIkCBuYiIiEgLCYJgr4cgu0SBuYiIiIhIBCgwFxEREWkhrVIvLTunwFxEREREJAIUmIuIiIi0ENWYty8F5iIiIiItRKUs7UuBuYiIiIhIBCgwFxERERGJAAXmIiIiIi1EpSztS4G5iIiIiEgEKDAXEREREYkABeYiIiIiIhGgwFxEREREJAIUmIuIiIiIRIACcxEREZEWoq4s7at7rwdQyBgTAz4G/IW19r15t/cC7wJeCaSAd1prfzVv+0HgD4BnAI8Ab7bWfrSZYxcRERERqVWkMubGmDjwW8Bzi2z+H8BNwHOAHwB+1hjzqvBxMeCDwBngeuCPgL8yxhxuxrhFREREROoVmYy5MWYB+FPgCHC+YNsg8P3A7dbae4F7jTG/CrwR+ABwG2CAW6y1y8B9xpjnAK8DfrZpP4SIiIiISI2ilDG/BngUuBZYKth2JdALfCrvtk8B1xtjuoCnAZ8Pg/L87Tft3nBFRERERBonMhlza+2HgQ8DGGMKN88BZ621ybzbTgIJYCbc/njBY04Ci7syWBEREZE9pkmg7adpgbkxpo/SgfLJgmx3oQHchM98uf/3ltneu9NxioiIiETZ8PAwAAMDA3s8Emm0ZmbMrwM+WWLb9wLvL/PYJNuD7Nz/18Lto0W2r+1siCIiIiLRFo+7SuTu7sgUPkiDNO03aq39FFDrNZfjwLgxJmGtTYe37cNlxc+G268seMw+4IkaX09EREREpKmiNPmznC8AaeDpebfdDNxrrc0C/wZcFXZvyd/+b00boYiIiIhIHVriGoi1ds0Y80fAu40x34PLhr8F10IR4J+Bh4H3G2PeCrwY16nldc0frYiIiIjIzrVKxhzgzcDngH8C3gO8zVr7FwDWWg94Ka5Dy73AdwMvs9Y+tDdDFRERERHZmUhmzK21h4rctga8NvxT7DEPAM/a3ZGJiIiIiOyOVsqYi4iIiIi0LQXmIiIiIiIRoMBcRERERCQCFJiLiIiIiESAAnMRERERkQhQYC4iIiIiEgEKzEVEREREIkCBuYiIiIhIBCgwFxERERGJAAXmIiIiIiIRoMBcRERERCQCFJiLiIiItKBYLLbXQ5AGU2AuIiIi0kIGBgZIJBK8/OUv3+uhSIN17/UARERERKR6vb29vPe976W7W2Fcu1HGXERERKTFKChvTwrMRUREREQiQIG5iIiIiEgEKDAXEREREYkABeYiIiIiIhGgwFxEREREJAIUmIuIiIiIRIACcxERERGRCFATTBERERFpS695zWt49NFH93oYVVNgLiIiIiJt6YUvfOFeD2FHVMoiIiIiIhIBCsxFRERERCJAgbmIiIiISAQoMBcRERERiQAF5iIiIiIiEaDAXEREREQkAhSYi4iIiIhEgAJzEREREZEIUGAuIiIiIhIBCsxFRERERCJAgbmIiIiISAQoMBcRERERiQAF5iIiIiIiEaDAXEREREQkAhSYi4iIiIhEgAJzEREREZEIUGAuIiIiIhIBCsxFRERERCJAgbmIiIiISAQoMBcRERERiQAF5iIiIiIiEaDAXEREREQkAhSYi4iIiIhEgAJzEREREZEIUGAuIiIiIhIBCsxFRERERCKge68HEBFdACdOnNjrcYiIiIhIG8uLN7sKtykwd+YA7rzzzr0eh4iIiIh0hjngwfwbFJg7nwNuAZ4AvD0ei4iIiIi0ry5cUP65wg2xIAiaPxwREREREdlCkz9FRERERCJAgbmIiIiISAQoMBcRERERiQAF5iIiIiIiEaDAXEREREQkAhSYi4iIiIhEgAJzEREREZEI0AJDe8QYEwM+BvyFtfa9ebdPAb8NPA9YBn7DWvvredsngN8Dng+cBX7BWvtHeduvBN4DXAncD/ygtXZbA/vdYozpBe4Ffsxa+/HwtkPA7wLPwC3i9DZr7Z/lPeYg8Afh9keAN1trP5q3/TbgN4GLgM8Cr7fWPlDLWKoZT979EsC/A39prX1rke23AP8MHLHWPlRpPJXGVs+4jDFxYAXoL7j7uLX2fIWxTOCOlWfk71djzFXAu4CrgAeAn7TWfixv+8dwx2m+l1lr/2+4/UeB/wRM4Y71N1prT5UbS5Gx1XI8zQPvBp4LnAF+2Vr77rztTT2eqhjPq4CfBw4An8Ltp4rjqWZ85cZmjHk/8NoiT/VNa+2R8D5N21dVjqfufVXjMXU58DvAtcDxIttr+tyt43tgP+6YeiZwDvgta+0787ZfRZn3brVq3FdlX3sP9lVd460wpqPAbwA3A6vA/wF+xlqbrPd7zRjzJuC/AqPAX+KO9dUyY/kW4JeBS3HH6K9aa/+wEa9XaR9HhTFmCLcPXob7LrwLeJO19ni4vRf3u34lkALeaa391SLPcxHwZWDYWpvNu70p8ZUy5nsgDKJ+C/dFXehvgItxb4DvAX48DHBy3g9M4t7sbwN+zxjz9PB5B4GPAv+G+wL5JPD/jDHDu/KDFDDG9AF/DlyWd1sv8HHcsfYM3Bv/t40xd4TbY8AHcQHL9cAfAX9ljDkcbt8PfAj4U+A64ATwwXAf7mgs1YynwM8WPr7g+d8LxMqNo9qxNWBcR4A+4BBuNbHcn6UKYxkHPgzMFNw+DfwT8A3gBtzx+tfGmOvy7nYZ8KqC1/to+PjXA7+EO0afhltR96OVfm8FY6jleIqHP08/7nj5r8CvG2OeG25v6vFUxXieA/wZ7li6Fvdh/4nwC6ZqNR5TP8rW391VuC/dd4aPb/Z7r9J46t5XNR5Tvbjf4RdxX8i/ArzfGHNjuL2mz906vwf+AljHfWb+KPCLxphXhs9bzXu3ohr3VdnXbva+qne8FcaUwB0XKeDpwJ3AHcA76v1eM8a8HHg78MPAbeFz/FqZsVwM/G24L67Cfe7+jjHm9ga9XqXjMSp+A7gVF3g/E/e5+3/zPrP+B3AT8BzgB4CfDU/2N4T76m9x36f5tzctvlLGvAnCs+o347Krn8a9OY4A5wvudy3uzPsSa60Nb/sJ3BfTb4bPcztwcXim++UwKP9h4F+B7wAywH+21vrGmB8HvjW8/b3sImPMMeB/sz1QfTHuS/YGa+3ZcMyX4D4g/y/uQ8AAt1hrl4H7wi/g1+EC0O8Hvpg7qzXGfB/uQ+XZuA/cnYylmvHknuMK4PXA10r8yP8NOAU8pcT2ourYT5XGdQx4xFr78A7GcjPwx7jgp9B3h7e/3lqbAe4PA5H/DLzaGDMCLAB3W2tPFHn8jwK/mbuaY4x5HfAY7ou1Yjaqjv30AtwXyHPD7fcbY56F+zD+B5p/PFUaz4/iMoC54PPN4WPuxF0Zq6jWfWWtXSLvxM0Y89u43+dvhzc1dV9VMZ669lUdx9Qx3Anvz4dXnx40xvwILgC4mxo+d40xC9T+PTCOO9n9IWvtfwD/YYz5O9zv5f+jwnu30n6qc19Veu2m7qsGjLecG3DZ5xustSvhY38ufO3/R33faz8GvMta+6Fw+w8CHzfGvCV8rULfAXzBWvvfw/8/EH7O3Ik7eaj59cKfo9w+bhhjzDOAHwL+p7X2Czt8bA/hyZG19l/D216Hu3rwFGPMo7j9cLu19l7gXmPMrwJvBD4Q3v8O4PdxV1YKNS2+UsZ8FxljnmaM+SvA4g7uLwPXAI/izrgKs5lHgHO5gz/0RWDOuMtxNwJPFFy6/RTuSx7ch/WnrbU+gLU2wJ0I3MTuexbwiSKvdcQNxZ7Nu+2LwHXhG+lpwOfDD6+cwp/pX3IbrLVruDKOcj9TqbFUMx6MMV3A+3Af3mcKnyDMprwGeEuZMex0bPWO6xjuONuJ54fP94oi244A94ZfVvnjyY37GJDEXaIt5gjwmdx/wt/b16n+WKx1Pz0b+ET+dmvtD1pr3xb+t9nHU6XxFO6nAPhShfFUO76Kx1SOMeYmXLbvx/Nubvp7r8J46t1Xte6n3O2vM8bEw7FdgtsXUNvnbj3fA+vAGvA9xpgeY4zBZYPvzXt8ufduNWrdV5Veu9n7qt7xlmOBFxUEygEwRh3fa+Hn/PX523FZ2m7g6hJj+QtcgJkvN5Z6X6/SPm6kbwADwD3GmH8wxmyUShpjDhljghJ/3oo7iXwp7njKCcK/x3BXu3pxv4ecTwHXh/sAXKD9c7gkQKGmxVfKmDeYcZewXgL8F9wlo/8NXG2t/XJ4l/txZ7C4z9MtTgIjxpjhvDf0wfDvKdyZ/+NFHrMY/nuO7cHZSdylrV1lrf3d3L8Lfq6TuDdwl7XWC287iDv2RqnuZyq3fSdjqWY8p3EB92lr7Z8YY34g/8Hhh/kf4jIq24L2SurYT2XHhQuUh40x/4LL0H4e+PGCD9PCsfxcOI5DRTafxH1Y5zuIOw5zr3ce+IBxtfaPAm+11n4k7/EbvyPjLiUu5j2+rDr201HgEWPM23H1yhdwdYS5WstmH0+VxlPstQ/ift9VqfOYyvkZ4K+stV/Ju20v3nvlxlPXvqp1P1lrHzbG/DSuNOtXgC7gF621/xDed8efu9baD1Pj94C19iFjzA/jan7fGI7nT+xm3XWl925FdRxTlV67qfuqAeMtyVr7JHlXjsLPuDeGt9XzvTaGK6PY2G6tzRpjzlDivRdeOdlgjJnFlRn+YgNe7zjl9/FDxcZUC2vtE8DLjTFHcMHxXxpjvgn8T9zJx1yJh65Ya9PA3xfc/qO47+kv4ILus9baZN72k0ACV8r5hLX2+wGMMbcWeY2mxVfKmDfeM3GXyB4HDlprvy8vKK/kblyA87vGmOHwEt5bw20J3JlkquAxKSARnhCU2t670x+igT6KG9d/N8b0hZcR3xxuK/cz5cbc6J+p7HiMMU/BnVT9YInH/xTwqLX2z2t8/d0a16XAOO54eSkum/0JY8xojeP5S+AaY8wbw6zc04Dvw/3Ocq83hKtbfAHwEeDD4WVgcJcGf9IYc6VxtZi/gPvwS1CfSsfTMO5qxhwu4/obuFrLO8L7NPV4qmI8HwB+wBjzLGNMt3G1+ddS/36qZmwAGDdJ7YWEtdx5mr2vKo1nt/ZVpfdeN+5k9w9xVy3fgKuxfXl4n0bvp0rfA+CuwH4Ul637DuCFxpgfC7dVeu/Wo9LvsNJrN3tf1TvenXgnLsP8X6nve20g7/+lHl+ScbXQf42LQXKTzOt5vWqOx4ay1n7DWvujwH5cCfAfA6+y1p4o8WdbeY8x5hW478z/EgbjpfYBVHf8NS2+UmDeePfhDqKXAv/LuLqyqlhrU7iyghtw2cgv40oNwGXakmw/CHqB9fCySqntazv7ERonzCp8Gy44WcVNtPmdcHO5nyk35pLbjTG3GGNW8v78dJ3jWcZ9+b7DFumyYoy5DNdl5Icqvc5O1TOu0C3Addbaf7LWfhb4TqAHeKkx5s6C/XRnFeO5DzfJ5+2438Ef4+oJL4R3+Ulgv7X2T621X7SuO8xHcBNqCB/3Sdzl0lVcYPORvMfXpIrjKYt777zBWnuvtfYPcPV/ud9ZM4+nasbzXtzv9h9wH/KvxE0Uq2s/VTm2nG8DHrDW3l3wFM3eV5XGsyv7qopxvQZXKvIjeb/Dd+Im2EH5/XSgYD+9p4rxlP0eMK67xg8D32ut/Zy19i+AnwDeGmaGK713a1ZpX1Xx2k3dVw0Yb0XGmJgx5jeBHwFeba39armfM/x3ue3JvP9v226M+emC/XRL3lhGgb/DlZ+8OCxZqev1qohLdoUx5gbc3JEfwCU7Hyr4uUt+/hg3ofPPgV+z1v6v8OZS+wCqi5GaFl8pMG8wa+2T1trX4i5h34eb4f1FY8xrw8xhpcf/u7X2KcA8sA9XA+XjanmPh7fl28fmRIVK2/dEeMl3AXdZbD+u1vh0eJZbz890D+4yUu5PxQ/ycuMBJnCTXN6ee8PjMlI/bYz5KO4Dfgw3yWcFV2cH8NVqgt1dHBfW2qTNa6UVZgi+GT7fh9i6nz5U5Xj+NHztRVyG7hzhZUtrrWe3t2G8P3w9rLVr1to7cZeLZ6y134nLGj9U3d4oO65yx9PjwNfzLluDu/x4IPx3046nasZjrfWttW8GRoB91trnh/9+qJrXrnNsOS/EZdgKNXtflR3Pbu6rCuO6HviyDWtLQ/figh8ov58eZ+t++vkqx1Pue+B63IlLfuu8e3Hvtcnw8SXfu/Wq9Dus8NrN3lf1jrcs48pX3oc70f4Oa+0Hq/g5K20/gwsCN7aHV20mw+3vYet+uie8zxRuXsAR4FZr7YN5z13P61Xcx41i3ByOlxljPoWreV8GnmqtfRmunvuqEn/ek/ccr8d1b/ota+1P5D39cWC8IA7bhzvJz5+DUErT4ivVmO8Sa+1jwFuMMb+Ie9P+Urjpj0o9xmy2rnu5tfZkeNtLgH+31l4wxvwbsGCMOZSXOb0ZN1GD8O+fNcbErLWBceUtz8DVRe6J8NLh7+K6UjwR3vYSXH9RcGP+aWPMYN4XTeHP9My85xvAXS58u7V2Hdd3tlHjOY7L7Ob7AK7jza/gJl39Wd62A8A/Ai9icyJYTeoZV/gh+jButnhudvlQ+JivhXWBxTqvlBvPrbg+tt9G+METjucT4b//Ejhlrf3hvIddjTsZxRjzK8B/2LCW2hizGG7/3p2Mo8i4Kh1PnwFeZIzpsZsTuo6x+UXbzOOp4njC8oNBa+07gGQ4nm/BdYuoSxVjy82JuQFXw1mo2fuq7Hh2a19VMa7HcZMh810K5AKfkp+71vVA3ul+qvQ98DhwxBjTG2Yzc+NZBp6s9N6tR6V9VcVrN3tf1TveSn4Nd3Xy5dbav827vZ7vNd8Y87nw/rka9ptwV98+H55QbAkkw2Dzb3E1388sCMrrer1K+7jK/VStW3AnOr8LvCL3euDq3qlwfBhjXobrqvJL1tqfKdj8BSCNa215V3jbzbjJv9kqxta0+EqB+S6zrgXYLxtj3omrAS5333PhG+bXjJtlfA0ua/CqcPs3jFvU5Y+NMW/E1VfeiWs5CK5e7peBdxlj3o1rDTRM2Apoj3wTl4V4R3hp8oW4SXA3h9v/GRdUvj/8mV+Mm/38unD7+4D/Yoz5GVwv1Z/DnaX/Y6PHU+yNb4xJ4iaMHA9vOpu3LffPh+3W2fdNH1d4XLwj/NI+C7wD90Xz4RrHY4EXGNer9kO4gPppuFaNhLf9fpjZ+BzwXbjfaa4G/jjwC8aY+3CXkP8A+GB4mbcelY6nD+DeM+81xrwDF+R9L67sAZp4PFU5nm8AfxJ+MT6MC0i/gSv7qVelsYGbxDUMFPu9NHtfVRrPbu2rSuP6U9x8iV/HTbi8ElfK9Z/D7Q393K30PYB7T/8KrlTyv+GywL+KW/QlMMZUeu/Wo9K+qvTazd5X9Y63JOPq0X8MN+/oHmNMfja13u+1dwN/YIz5Unj7u4H32eKtEsF1L7oWN99nNW8saes60tT1ehX2cSN9CVciWernLClMRv0+7gTlXQW/j7PW2jVjzB8B7zbGfA8u2/0W3DFYjabFVyplaRJrbTr/7K+M78BddvsSru7t+23YWzT03bg6r7txb47XW2s/E77GBdzM46fjMrjPwLVzqjdorFmY0bkDd/LwFdzBfId1fUQJL/G/FDcx8F7cz/ey3BWB8O+X42oE7wFmgZcWXFZu2Hj2SgPG9SZcgPJ/cMcGwAuqzAQUG88TuNKdN4TjeTbwnPBKENbaP8Z9qL0NV3P4IuD5eZma38H1VP4QLvv0BYqv6rjTcVU6npZxi0fk3kO/iKsN/nC4/SGaeDxVMZ4P4U6i/hduNb414FutmzNSlyqPqdnw722XcvfovVduPLuyr6r4HT6Ey8xfhzuOfxn4aWvt+8Ltu/G5W/J7IEz2PBuX6LkbV3v/PtzaChXfu/WoYl9V+txo9r6qa7wVfFv49y/hkiD5f3Kt+2r6XguvfL4dlzn+eHif3IlgMa/EJVo/XjCO3H6o9/UqxSUNYa09V0tQHroVd8Xgdrb/PnJXC96MSyT9E6785W3WzdGoZmxNi69iQVD357+IiIiIiNRJGXMRERERkQhQYC4iIiIiEgEKzEVEREREIkCBuYiIiIhIBCgwFxERERGJAAXmIiIiIiIRoMBcRETqYowJjDHPqfGxbzfG3NXgIYmItCSt/CkiIvWao8iiQCIisjMKzEVEpC7W2hN7PQYRkXagwFxEpMMYYw4B3wS+C/gVYAj4U+DHrbUZY8zNwDuBpwLfAH7ZWvsn4WPfj1ty/ApgP2658y8Bz7XWftwY0we8FbgTmMAtf/1Ga+3D4eOPAb8PXAN8Gnhg939iEZHWoBpzEZHO9fPAq4E7wj9vN8bsAz4C/BkuMH8b8C5jzO15j/uu8PYXAl8teM73AK8Avht4Gi4B9CFjTJcxphf4f7iTgmuAvwG+fzd+MBGRVqSMuYhI5/pJa+0nAYwxPwf8TyANfMJa+5vhfR4wxlwC/Bjw4fC2z1tr/yb3JMaY3N/jwGuAF1trPxHedifwKPCC8O7TwA9Za1eArxljng1M7dpPKCLSQhSYi4h0rn/N+/c9uNKTpwO3GGNW8rZ1A0/m/f+hEs/3FNyV2LtzN1hrzxpjLHAprgTmwTAoz3/dFyAiIgrMRUQ6WDbv313h32ngz4FfLLivl/fvZInnWy9xe1f4x8cF5/kylYcpItIZVGMuItK5rsr793XASdxEzouttQ/k/gDPB15fxfM9iAv2b8zdYIyZBC4GLPAV4KKw5CXn6rp+AhGRNqKMuYhI5/p1Y8z3AaPAfwN+B/hj4D8ZY34JeB9wJfCrwFsqPZm1dtUY8x7gt4wxbwBO47q+HAf+Dpd1fxh4nzHmZ3CTQ78N+LdG/2AiIq1IGXMRkc71AeBvw7/fB7wjbGv4YuA5uAz3rwG/YK393Sqf8yeAjwF/iathTwHPttYmrbUZ4EXACHAv8Abg3Y37cUREWlssCIK9HoOIiDRRXh/zi8NSFRERiQBlzEVEREREIkCBuYiIiIhIBKiURUREREQkApQxFxERERGJAAXmIiIiIiIRoMBcRERERCQCFJiLiIiIiESAAnMRERERkQhQYC4iIiIiEgH/PzKwnikWTE69AAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 864x864 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "bins = dataset.TABULA_AGE_BINS['harmonized']\n",
    "labels = utils.generate_labels(bins)\n",
    "\n",
    "predictor = regressor_fts_importance_fr\n",
    "df_res = predictor.individual_prediction_error()\n",
    "df_res['period'] = pd.cut(predictor.y_test[dataset.AGE_ATTRIBUTE], bins, labels=labels, right=False) # not metric\n",
    "df_res = df_res.join(predictor.aux_vars_test, how=\"inner\")\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    sns.violinplot(data=df_res, x='period', y='error', hue='type', inner='quartile', split=True, palette='Set2', ax=ax)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 423,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Boolean inputs to the `inclusive` argument are deprecated in favour of `both` or `neither`.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "Boolean inputs to the `inclusive` argument are deprecated in favour of `both` or `neither`.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "Boolean inputs to the `inclusive` argument are deprecated in favour of `both` or `neither`.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "Boolean inputs to the `inclusive` argument are deprecated in favour of `both` or `neither`.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "Boolean inputs to the `inclusive` argument are deprecated in favour of `both` or `neither`.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    }
   ],
   "source": [
    "y_preds = []\n",
    "y_tests = []\n",
    "# predictors = {'Netherlands': regressor_fts_importance_nl, 'France': regressor_fts_importance_fr, 'Spain': regressor_fts_importance_es}\n",
    "predictors = {'Country cv': regressor_country_cv, 'City cv': regressor_nl_city_cv, 'Neighborhood cv': regressor_fts_importance_nl, 'Block cv': regressor_nl_block_cv, 'Random cv': regressor_nl_random_cv}\n",
    "# predictors = {'City cv': regressor_nl_city_cv, 'Neighborhood cv': regressor_fts_importance_nl, 'Block cv': regressor_nl_block_cv, 'Random cv': regressor_nl_random_cv}\n",
    "\n",
    "for c, p in predictors.items():\n",
    "    y_pred = p.y_predict[p.y_predict[dataset.AGE_ATTRIBUTE].between(1900, 2017, inclusive=False)]\n",
    "    y_test = p.y_test.loc[y_pred.index]\n",
    "    y_pred['country'] = c\n",
    "    y_test['country'] = c\n",
    "    y_preds.append(y_pred)\n",
    "    y_tests.append(y_test)\n",
    "\n",
    "# y_test = pd.concat(y_tests)\n",
    "# y_pred = pd.concat(y_preds)\n",
    "\n",
    "\n",
    "# visualizations.plot_relative_grid(y_test, y_pred, bin_size=1, figsize=(8, 8))\n",
    "# plt.savefig('../figures/error-gridplot-combined.svg')\n",
    "\n",
    "for i, c in enumerate(predictors.keys()):\n",
    "    visualizations.plot_relative_grid(y_tests[i], y_preds[i], bin_size=2, figsize=(8, 8))\n",
    "    plt.savefig(f\"../figures/error-gridplot-{c.replace(' ', '-')}.svg\")\n",
    "\n",
    "# sns.jointplot(y=y_test['age'], x=y_pred['age'], kind=\"hex\")\n",
    "\n",
    "\n",
    "\n",
    "# visualizations.plot_grid(y_test, y_pred)\n",
    "# plt.savefig('../figures/feature-distributions.svg')\n",
    "# import matplotlib\n",
    "# colors = ['plum', 'thistle', 'lavenderblush', 'paleturquoise', 'mediumturquoise']\n",
    "# matplotlib.colors.LinearSegmentedColormap.from_list(\"\", ['thistle', 'paleturquoise', 'steelblue'])\n",
    "# matplotlib.colors.LinearSegmentedColormap.from_list(\"\", [\"red\",\"violet\",\"blue\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# def mean_bin_age(s, bins):\n",
    "#     cat_to_metric = {idx: (bins[idx] + bins[idx+1]) / 2 for idx in range(len(bins) - 1)}\n",
    "#     return s.map(cat_to_metric)\n",
    "\n",
    "# bins = dataset.TABULA_AGE_BINS['harmonized']\n",
    "# labels = utils.generate_labels(bins)\n",
    "\n",
    "# error_dfs = []\n",
    "# for predictor, city_boundaries, gadm_level, dissolve_level in [(regressor_fts_importance_fr, gadm_france, 4, 2), (regressor_fts_importance_es, gadm_spain, 4, 2), (regressor_fts_importance_nl, gadm_netherlands, 2, 1)]:\n",
    "#     y = predictor.individual_prediction_error()\n",
    "#     y['period'] = predictor.y_test[dataset.AGE_ATTRIBUTE]\n",
    "#     # y['period'] = pd.cut(predictor.y_test[dataset.AGE_ATTRIBUTE], bins, labels=labels, right=False) # not metric\n",
    "#     # y['period'] = utils.custom_round(predictor.y_test[dataset.AGE_ATTRIBUTE], base=5) # moving average\n",
    "\n",
    "#     y = pd.concat([y, predictor.aux_vars_test], axis=1, join=\"inner\")\n",
    "#     y['error'] = y['error'].abs()\n",
    "#     error_by_city = y #.groupby(['city', 'period'])['error'].mean().reset_index()\n",
    "#     error_by_city = error_by_city[['city', 'period', 'error']].merge(city_boundaries[[f'NAME_{gadm_level}', 'NAME_0', 'DEGURBA']], right_on=f'NAME_{gadm_level}', left_on='city', how='left')\n",
    "#     error_dfs.append(error_by_city)\n",
    "\n",
    "# df = pd.concat(error_dfs)\n",
    "\n",
    "linestyles = ['-', '--', ':', '-.']\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    # for i, (country, group) in enumerate(df.groupby('DEGURBA')):  # plot average cross countries\n",
    "    for i, (country, group) in enumerate(df.groupby('NAME_0')):  # plot average across city types\n",
    "        error_df = group.groupby('period')['error'].agg(['mean', 'std']).reset_index()\n",
    "        error_df.plot(x='period', y='mean', kind='line', ax=ax, xlabel='Construction period', ylabel='MAE', xlim=(1900, 2020), ylim=(0, 50), label=country, style=linestyles[i])\n",
    "        ax.set_xticks([1920,1940,1960,1980,2000], labels=['1920', '1940', '1960', '1980', '2000'])\n",
    "        ax.legend(loc='upper center')\n",
    "        # plot either std around mean or min max of DEBURBA types\n",
    "        # groupby DEBURBA to get min max of deburga instead of min max of all buildings in this period\n",
    "        error_degurba_df = group.groupby(['period', 'DEGURBA'])['error'].mean().reset_index().groupby('period')['error'].agg(['max', 'min', 'std']).reset_index()\n",
    "\n",
    "        ax.fill_between(\n",
    "            error_df['period'],\n",
    "            # error_degurba_df['min'],\n",
    "            # error_degurba_df['max'],\n",
    "            error_df['mean'] - error_df['std'],\n",
    "            error_df['mean'] + error_df['std'],\n",
    "            alpha=0.1,\n",
    "        )\n",
    "\n",
    "        plt.savefig(f'../figures/error-over-years.svg')\n",
    "        # for i, (degurba, d_group) in enumerate(group.groupby('DEGURBA')):\n",
    "        #     d_group.plot(x='period', y='mean', kind='line', style=linestyles[i], ax=ax, ylim=(0, 50), label=f'{country} {degurba}')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 538,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "21.89838216205763"
      ]
     },
     "execution_count": 538,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df['period'].isin(['1960-1969','1970-1979','1980-1989'])]['error'].mean()\n",
    "df[df['period'].isin(['1900-1944'])]['error'].mean()\n",
    "df[df['period'].isin(['2000-2009', '>=2010'])]['error'].mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 534,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>Country</th>\n",
       "      <th>France</th>\n",
       "      <th>Netherlands</th>\n",
       "      <th>Spain</th>\n",
       "      <th>All</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Construction period</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1900-1944</th>\n",
       "      <td>33.821938</td>\n",
       "      <td>27.844218</td>\n",
       "      <td>37.985481</td>\n",
       "      <td>33.217212</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1945-1959</th>\n",
       "      <td>12.024916</td>\n",
       "      <td>12.591786</td>\n",
       "      <td>15.417421</td>\n",
       "      <td>13.344708</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1960-1969</th>\n",
       "      <td>11.114906</td>\n",
       "      <td>7.770391</td>\n",
       "      <td>11.849584</td>\n",
       "      <td>10.244960</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1970-1979</th>\n",
       "      <td>9.658246</td>\n",
       "      <td>6.558260</td>\n",
       "      <td>11.912170</td>\n",
       "      <td>9.376225</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1980-1989</th>\n",
       "      <td>9.910322</td>\n",
       "      <td>8.278921</td>\n",
       "      <td>12.137087</td>\n",
       "      <td>10.108776</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1990-1999</th>\n",
       "      <td>15.287704</td>\n",
       "      <td>12.572493</td>\n",
       "      <td>14.132315</td>\n",
       "      <td>13.997504</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2000-2009</th>\n",
       "      <td>20.766391</td>\n",
       "      <td>21.065634</td>\n",
       "      <td>18.134643</td>\n",
       "      <td>19.988889</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>&gt;=2010</th>\n",
       "      <td>27.957152</td>\n",
       "      <td>29.401329</td>\n",
       "      <td>25.430972</td>\n",
       "      <td>27.596484</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>17.567697</td>\n",
       "      <td>15.760379</td>\n",
       "      <td>18.374959</td>\n",
       "      <td>17.234345</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Country                 France  Netherlands      Spain        All\n",
       "Construction period                                              \n",
       "1900-1944            33.821938    27.844218  37.985481  33.217212\n",
       "1945-1959            12.024916    12.591786  15.417421  13.344708\n",
       "1960-1969            11.114906     7.770391  11.849584  10.244960\n",
       "1970-1979             9.658246     6.558260  11.912170   9.376225\n",
       "1980-1989             9.910322     8.278921  12.137087  10.108776\n",
       "1990-1999            15.287704    12.572493  14.132315  13.997504\n",
       "2000-2009            20.766391    21.065634  18.134643  19.988889\n",
       ">=2010               27.957152    29.401329  25.430972  27.596484\n",
       "All                  17.567697    15.760379  18.374959  17.234345"
      ]
     },
     "execution_count": 534,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "periods_df = df.groupby(['NAME_0', 'period'])['error'].mean().reset_index()\n",
    "periods_df = periods_df.rename(columns={'NAME_0': 'Country', 'period': 'Construction period', 'error': 'MAE'})\n",
    "pivot_table = periods_df[['MAE', 'Country', 'Construction period']].pivot_table(values=['MAE'], index=['Construction period'], columns=['Country'], margins=True)\n",
    "\n",
    "pivot_table.columns = pivot_table.columns.droplevel(0) # drop \"MAE\" multi-index column\n",
    "# pivot_table.to_latex('../figures/prediction-error-by-periods.tex', index=True, float_format='%.2f', caption='Prediction error across construction periods', label=table:error-by-periods)\n",
    "pivot_table\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure: Pivot table of construction period, city type and country"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean error: 0.004688806440203154\n",
      "Mean error: -0.03462956125813288\n",
      "Mean error: 0.013162852881266677\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-08-05 12:06:19,682 | WARNING : Building type information missing. Considering all buildings as residential.\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "\n",
    "dfs = []\n",
    "error_dfs = []\n",
    "# bins = utils.generate_bins((1900, 2020, 10))\n",
    "bins = dataset.TABULA_AGE_BINS['harmonized']\n",
    "labels = utils.generate_labels(bins)\n",
    "for predictor, city_boundaries, gadm_level, dissolve_level in [(regressor_fts_importance_fr, gadm_france, 4, 2), (regressor_fts_importance_es, gadm_spain, 4, 2), (regressor_fts_importance_nl, gadm_netherlands, 2, 1)]:\n",
    "# for predictor, city_boundaries, gadm_level, dissolve_level in [(regressor_fts_importance_fr, gadm_france, 4, 2)]:\n",
    "    df = predictor.individual_prediction_error()\n",
    "    df['error'] = df['error']\n",
    "    df['y_test'] = predictor.y_test[dataset.AGE_ATTRIBUTE]\n",
    "    df['y_predict'] = predictor.y_predict[dataset.AGE_ATTRIBUTE]\n",
    "    df['period'] = pd.cut(predictor.y_test[dataset.AGE_ATTRIBUTE], bins, labels=labels, right=False)\n",
    "    # df['period'].replace(dict(zip(range(1,8), labels[1:])), inplace=True)\n",
    "    print(f'Mean error: {df[\"error\"].mean()}')\n",
    "    df = df.join(predictor.aux_vars_test, how=\"inner\")\n",
    "    df = df.join(predictor.X_test[['FootprintArea']], how=\"inner\")\n",
    "    df = preparation.add_residential_type_column(df)\n",
    "\n",
    "\n",
    "    df = df.merge(city_boundaries[[f'NAME_{gadm_level}', 'NAME_0', 'DEGURBA']], right_on=f'NAME_{gadm_level}', left_on='city', how='left').set_index('id')\n",
    "\n",
    "    error_by_city = df.groupby(['NAME_0', 'DEGURBA', 'type', 'period'], dropna=False).apply(lambda x: pd.Series({\n",
    "            'MAE'       : x['error'].abs().mean(),\n",
    "            'MAE-metrics': metrics.mean_absolute_error(x['y_test'], x['y_predict']),\n",
    "            'n'         : len(x.index),\n",
    "            'R2'        : metrics.r2_score(x['y_test'], x['y_predict']) # doesn't make sense for grouping on period\n",
    "        })).reset_index()\n",
    "\n",
    "    # error_by_city = geometry.to_gdf(error_by_city)\n",
    "    error_dfs.append(error_by_city)\n",
    "    dfs.append(df)\n",
    "\n",
    "\n",
    "df = pd.concat(dfs)\n",
    "error_df = pd.concat(error_dfs)\n",
    "\n",
    "# old approach\n",
    "# error_df[error_df['size'] > 10000].sort_values(by=['mean'])\n",
    "# error_df.groupby('NAME_0').apply(lambda x: np.average(x['mean'], weights=x['size']))\n",
    "# df = error_df.groupby(['NAME_0', 'DEGURBA', 'type', 'period'])['error'].agg(['count', 'mean', 'std']).reset_index()\n",
    "\n",
    "# column renaming\n",
    "error_df['DEGURBA'].replace({1.0: '1 (city)', 2.0: '2 (town)', 3.0: '3 (rural)'}, inplace=True)\n",
    "error_df = error_df.rename(columns={'NAME_0': 'Country', 'period': 'Construction period'})\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 254,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Netherlands</th>\n",
       "      <th colspan=\"3\" halign=\"left\">France</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Spain</th>\n",
       "      <th>All</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DEGURBA</th>\n",
       "      <th>1 (city)</th>\n",
       "      <th>2 (town)</th>\n",
       "      <th>3 (rural)</th>\n",
       "      <th>1 (city)</th>\n",
       "      <th>2 (town)</th>\n",
       "      <th>3 (rural)</th>\n",
       "      <th>1 (city)</th>\n",
       "      <th>2 (town)</th>\n",
       "      <th>3 (rural)</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Construction period</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1900-1944</th>\n",
       "      <td>23.060449</td>\n",
       "      <td>33.382057</td>\n",
       "      <td>34.573950</td>\n",
       "      <td>30.415801</td>\n",
       "      <td>36.098210</td>\n",
       "      <td>40.432567</td>\n",
       "      <td>42.560626</td>\n",
       "      <td>46.391679</td>\n",
       "      <td>38.906994</td>\n",
       "      <td>37.877814</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1945-1969</th>\n",
       "      <td>9.516255</td>\n",
       "      <td>9.771991</td>\n",
       "      <td>10.421881</td>\n",
       "      <td>11.240884</td>\n",
       "      <td>11.509832</td>\n",
       "      <td>12.865186</td>\n",
       "      <td>16.600570</td>\n",
       "      <td>18.237625</td>\n",
       "      <td>14.092643</td>\n",
       "      <td>13.492874</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1970-1979</th>\n",
       "      <td>6.771781</td>\n",
       "      <td>6.251914</td>\n",
       "      <td>6.995713</td>\n",
       "      <td>12.099940</td>\n",
       "      <td>10.184193</td>\n",
       "      <td>10.477131</td>\n",
       "      <td>10.848363</td>\n",
       "      <td>13.648571</td>\n",
       "      <td>13.457767</td>\n",
       "      <td>11.055586</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1980-1989</th>\n",
       "      <td>8.255361</td>\n",
       "      <td>7.958356</td>\n",
       "      <td>9.375220</td>\n",
       "      <td>17.229033</td>\n",
       "      <td>12.224253</td>\n",
       "      <td>12.697736</td>\n",
       "      <td>9.154319</td>\n",
       "      <td>10.661819</td>\n",
       "      <td>15.970319</td>\n",
       "      <td>12.352446</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1990-1999</th>\n",
       "      <td>12.272930</td>\n",
       "      <td>12.426067</td>\n",
       "      <td>13.721617</td>\n",
       "      <td>24.122348</td>\n",
       "      <td>17.657456</td>\n",
       "      <td>18.559442</td>\n",
       "      <td>9.443355</td>\n",
       "      <td>10.537594</td>\n",
       "      <td>18.736885</td>\n",
       "      <td>15.980565</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2000-2009</th>\n",
       "      <td>20.459807</td>\n",
       "      <td>21.094188</td>\n",
       "      <td>22.623677</td>\n",
       "      <td>30.446464</td>\n",
       "      <td>24.399581</td>\n",
       "      <td>25.383097</td>\n",
       "      <td>15.548013</td>\n",
       "      <td>14.566305</td>\n",
       "      <td>21.614763</td>\n",
       "      <td>21.907254</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>&gt;=2010</th>\n",
       "      <td>29.956646</td>\n",
       "      <td>28.854132</td>\n",
       "      <td>29.589343</td>\n",
       "      <td>39.001135</td>\n",
       "      <td>31.500865</td>\n",
       "      <td>32.247837</td>\n",
       "      <td>22.857034</td>\n",
       "      <td>23.707597</td>\n",
       "      <td>41.495764</td>\n",
       "      <td>31.468134</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>15.756176</td>\n",
       "      <td>17.105529</td>\n",
       "      <td>18.185914</td>\n",
       "      <td>23.507944</td>\n",
       "      <td>20.510627</td>\n",
       "      <td>21.808999</td>\n",
       "      <td>18.144612</td>\n",
       "      <td>19.678742</td>\n",
       "      <td>23.467876</td>\n",
       "      <td>20.590668</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Country             Netherlands                           France             \\\n",
       "DEGURBA                1 (city)   2 (town)  3 (rural)   1 (city)   2 (town)   \n",
       "Construction period                                                           \n",
       "1900-1944             23.060449  33.382057  34.573950  30.415801  36.098210   \n",
       "1945-1969              9.516255   9.771991  10.421881  11.240884  11.509832   \n",
       "1970-1979              6.771781   6.251914   6.995713  12.099940  10.184193   \n",
       "1980-1989              8.255361   7.958356   9.375220  17.229033  12.224253   \n",
       "1990-1999             12.272930  12.426067  13.721617  24.122348  17.657456   \n",
       "2000-2009             20.459807  21.094188  22.623677  30.446464  24.399581   \n",
       ">=2010                29.956646  28.854132  29.589343  39.001135  31.500865   \n",
       "All                   15.756176  17.105529  18.185914  23.507944  20.510627   \n",
       "\n",
       "Country                             Spain                              All  \n",
       "DEGURBA              3 (rural)   1 (city)   2 (town)  3 (rural)             \n",
       "Construction period                                                         \n",
       "1900-1944            40.432567  42.560626  46.391679  38.906994  37.877814  \n",
       "1945-1969            12.865186  16.600570  18.237625  14.092643  13.492874  \n",
       "1970-1979            10.477131  10.848363  13.648571  13.457767  11.055586  \n",
       "1980-1989            12.697736   9.154319  10.661819  15.970319  12.352446  \n",
       "1990-1999            18.559442   9.443355  10.537594  18.736885  15.980565  \n",
       "2000-2009            25.383097  15.548013  14.566305  21.614763  21.907254  \n",
       ">=2010               32.247837  22.857034  23.707597  41.495764  31.468134  \n",
       "All                  21.808999  18.144612  19.678742  23.467876  20.590668  "
      ]
     },
     "execution_count": 254,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pivot_table = error_df[['MAE', 'DEGURBA', 'Country', 'Construction period']].pivot_table(values=['MAE'], index=['Construction period'], columns=['Country', 'DEGURBA'], margins=True)\n",
    "# error_df[['MAE', 'DEGURBA', 'Country', 'Construction period']].pivot_table(values=['MAE'], index=['Country', 'Construction period'], columns=['DEGURBA'], margins=True)\n",
    "\n",
    "pivot_table.columns = pivot_table.columns.droplevel(0) # drop \"MAE\" multi-index column\n",
    "new_cols = pivot_table.columns.reindex(['Netherlands', 'France', 'Spain', 'All'], level=0)\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "\n",
    "pivot_table.to_latex('../figures/prediction-error-by-all.tex', index=True, float_format='%.2f', caption='Prediction error by city type, construction period, and country')\n",
    "pivot_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th>DEGURBA</th>\n",
       "      <th colspan=\"2\" halign=\"left\">1 (city)</th>\n",
       "      <th colspan=\"2\" halign=\"left\">2 (town)</th>\n",
       "      <th colspan=\"2\" halign=\"left\">3 (rural)</th>\n",
       "      <th>All</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>type</th>\n",
       "      <th>non-residential</th>\n",
       "      <th>residential</th>\n",
       "      <th>non-residential</th>\n",
       "      <th>residential</th>\n",
       "      <th>non-residential</th>\n",
       "      <th>residential</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Construction period</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1900-1944</th>\n",
       "      <td>24.732873</td>\n",
       "      <td>23.419730</td>\n",
       "      <td>30.636411</td>\n",
       "      <td>30.286030</td>\n",
       "      <td>36.812788</td>\n",
       "      <td>37.187703</td>\n",
       "      <td>30.512589</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1945-1969</th>\n",
       "      <td>10.647000</td>\n",
       "      <td>10.140225</td>\n",
       "      <td>10.684128</td>\n",
       "      <td>10.929941</td>\n",
       "      <td>11.424997</td>\n",
       "      <td>12.179645</td>\n",
       "      <td>11.000989</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1970-1979</th>\n",
       "      <td>13.790596</td>\n",
       "      <td>9.109822</td>\n",
       "      <td>10.576323</td>\n",
       "      <td>8.414397</td>\n",
       "      <td>10.492754</td>\n",
       "      <td>9.826589</td>\n",
       "      <td>10.368414</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1980-1989</th>\n",
       "      <td>20.644176</td>\n",
       "      <td>12.494623</td>\n",
       "      <td>13.451253</td>\n",
       "      <td>8.153282</td>\n",
       "      <td>12.963348</td>\n",
       "      <td>9.305366</td>\n",
       "      <td>12.835341</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1990-1999</th>\n",
       "      <td>27.837347</td>\n",
       "      <td>19.909247</td>\n",
       "      <td>21.480220</td>\n",
       "      <td>12.099535</td>\n",
       "      <td>22.145698</td>\n",
       "      <td>13.111005</td>\n",
       "      <td>19.430508</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2000-2009</th>\n",
       "      <td>30.166346</td>\n",
       "      <td>29.050888</td>\n",
       "      <td>25.950293</td>\n",
       "      <td>20.115064</td>\n",
       "      <td>27.778210</td>\n",
       "      <td>18.569269</td>\n",
       "      <td>25.271678</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>&gt;=2010</th>\n",
       "      <td>39.596078</td>\n",
       "      <td>38.356706</td>\n",
       "      <td>32.388897</td>\n",
       "      <td>27.698244</td>\n",
       "      <td>35.675800</td>\n",
       "      <td>25.411278</td>\n",
       "      <td>33.187834</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>23.916345</td>\n",
       "      <td>20.354463</td>\n",
       "      <td>20.738218</td>\n",
       "      <td>16.813785</td>\n",
       "      <td>22.470514</td>\n",
       "      <td>17.941551</td>\n",
       "      <td>20.372479</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "DEGURBA                    1 (city)                    2 (town)              \\\n",
       "type                non-residential residential non-residential residential   \n",
       "Construction period                                                           \n",
       "1900-1944                 24.732873   23.419730       30.636411   30.286030   \n",
       "1945-1969                 10.647000   10.140225       10.684128   10.929941   \n",
       "1970-1979                 13.790596    9.109822       10.576323    8.414397   \n",
       "1980-1989                 20.644176   12.494623       13.451253    8.153282   \n",
       "1990-1999                 27.837347   19.909247       21.480220   12.099535   \n",
       "2000-2009                 30.166346   29.050888       25.950293   20.115064   \n",
       ">=2010                    39.596078   38.356706       32.388897   27.698244   \n",
       "All                       23.916345   20.354463       20.738218   16.813785   \n",
       "\n",
       "DEGURBA                   3 (rural)                    All  \n",
       "type                non-residential residential             \n",
       "Construction period                                         \n",
       "1900-1944                 36.812788   37.187703  30.512589  \n",
       "1945-1969                 11.424997   12.179645  11.000989  \n",
       "1970-1979                 10.492754    9.826589  10.368414  \n",
       "1980-1989                 12.963348    9.305366  12.835341  \n",
       "1990-1999                 22.145698   13.111005  19.430508  \n",
       "2000-2009                 27.778210   18.569269  25.271678  \n",
       ">=2010                    35.675800   25.411278  33.187834  \n",
       "All                       22.470514   17.941551  20.372479  "
      ]
     },
     "execution_count": 182,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pivot_table = error_df[error_df['Country'] == 'France'][['MAE', 'DEGURBA', 'type', 'Construction period']].pivot_table(values=['MAE'], index=['Construction period'], columns=['DEGURBA', 'type'], margins=True)\n",
    "pivot_table.columns = pivot_table.columns.droplevel(0) # drop \"MAE\" multi-index column\n",
    "pivot_table.to_latex('../figures/prediction-error-type-FRA.tex', index=True, float_format='%.2f', caption='Prediction error in France by city type, construction period, and building usage')\n",
    "pivot_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "NAME_0  type           \n",
       "France  non-residential    22.484110\n",
       "        residential        17.200683\n",
       "Spain   non-residential    18.502431\n",
       "        residential        19.399889\n",
       "        unknown            26.338117\n",
       "Name: error, dtype: float64"
      ]
     },
     "execution_count": 182,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.groupby(['NAME_0', 'type'])['error'].apply(lambda x: x.abs().mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 280,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>non-residential</th>\n",
       "      <th>residential</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <th>DEGURBA</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">France</th>\n",
       "      <th>1.0</th>\n",
       "      <td>21.155374</td>\n",
       "      <td>17.436502</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2.0</th>\n",
       "      <td>20.632561</td>\n",
       "      <td>15.535117</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3.0</th>\n",
       "      <td>22.839069</td>\n",
       "      <td>17.484174</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>21.542335</td>\n",
       "      <td>16.818598</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">Spain</th>\n",
       "      <th>1.0</th>\n",
       "      <td>16.177632</td>\n",
       "      <td>15.565726</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2.0</th>\n",
       "      <td>15.568639</td>\n",
       "      <td>17.274368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3.0</th>\n",
       "      <td>22.339844</td>\n",
       "      <td>24.367939</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>18.028705</td>\n",
       "      <td>19.069344</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 non-residential  residential\n",
       "Country DEGURBA                              \n",
       "France  1.0            21.155374    17.436502\n",
       "        2.0            20.632561    15.535117\n",
       "        3.0            22.839069    17.484174\n",
       "        All            21.542335    16.818598\n",
       "Spain   1.0            16.177632    15.565726\n",
       "        2.0            15.568639    17.274368\n",
       "        3.0            22.339844    24.367939\n",
       "        All            18.028705    19.069344"
      ]
     },
     "execution_count": 280,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_country = df[df['NAME_0'].isin(['France', 'Spain'])]\n",
    "df_country = df_country.rename(columns={'NAME_0': 'Country'})\n",
    "# df_country['type'].fillna('unknown', inplace=True)\n",
    "df_country = df_country[df_country['type'] != 'unknown']\n",
    "\n",
    "df_country = df_country.groupby(['Country', 'DEGURBA', 'type'])['error'].apply(lambda x: x.abs().mean()).reset_index()\n",
    "pivot_table = df_country.pivot_table(values=['error'], index=['Country', 'DEGURBA'], columns=['type'], margins=False)\n",
    "pivot_table.columns = pivot_table.columns.droplevel(0) # drop \"MAE\" multi-index column\n",
    "pivot_table.columns = pivot_table.columns.rename(None) # drop Country cell\n",
    "pivot_table.loc[('France', 'All'),:] = df_country[df_country['Country'] == 'France'].groupby(['type'])['error'].apply(lambda x: x.abs().mean())\n",
    "pivot_table.loc[('Spain', 'All'),:] = df_country[df_country['Country'] == 'Spain'].groupby(['type'])['error'].apply(lambda x: x.abs().mean())\n",
    "pivot_table = pivot_table.sort_index()\n",
    "pivot_table.to_latex('../figures/prediction-error-type-DEGURBA.tex', index=True, float_format='%.2f', caption='Prediction error by building and settlement type in France and Spain', label='table:prediction-error-residential')\n",
    "pivot_table\n",
    "\n",
    "# df_country = df_country.groupby(['Country', 'period', 'type'])['error'].apply(lambda x: x.abs().mean()).reset_index()\n",
    "# pivot_table = df_country.pivot_table(values=['error'], index=['period'], columns=['Country', 'type'], margins=False)\n",
    "# pivot_table.columns = pivot_table.columns.droplevel(0) # drop \"MAE\" multi-index column\n",
    "# pivot_table.columns = pivot_table.columns.rename(None, level=0) # drop Country cell\n",
    "# pivot_table = pivot_table.sort_index()\n",
    "# pivot_table.loc['All',:] = df_country.groupby(['Country', 'type'])['error'].apply(lambda x: x.abs().mean())\n",
    "# pivot_table.to_latex('../figures/prediction-error-type-periods.tex', index=True, float_format='%.2f', caption='Prediction error by building type across construction periods in France and Spain', label='table:prediction-error-residential-periods')\n",
    "# pivot_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 287,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th colspan=\"3\" halign=\"left\">France</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Spain</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>metatype</th>\n",
       "      <th>type</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">residential</th>\n",
       "      <th>TH</th>\n",
       "      <td>18.092241</td>\n",
       "      <td>24.672725</td>\n",
       "      <td>0.431736</td>\n",
       "      <td>21.284345</td>\n",
       "      <td>28.612939</td>\n",
       "      <td>0.275659</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SFH</th>\n",
       "      <td>16.048118</td>\n",
       "      <td>21.978529</td>\n",
       "      <td>0.376297</td>\n",
       "      <td>16.545053</td>\n",
       "      <td>22.494812</td>\n",
       "      <td>0.257353</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MFH</th>\n",
       "      <td>20.716695</td>\n",
       "      <td>27.467044</td>\n",
       "      <td>0.309220</td>\n",
       "      <td>16.002098</td>\n",
       "      <td>22.288014</td>\n",
       "      <td>0.296023</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>AB</th>\n",
       "      <td>15.622602</td>\n",
       "      <td>20.550609</td>\n",
       "      <td>0.035450</td>\n",
       "      <td>10.933419</td>\n",
       "      <td>14.723366</td>\n",
       "      <td>0.396498</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">non-residential</th>\n",
       "      <th>others</th>\n",
       "      <td>21.866204</td>\n",
       "      <td>28.357705</td>\n",
       "      <td>0.276258</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>industrial</th>\n",
       "      <td>15.761084</td>\n",
       "      <td>22.346697</td>\n",
       "      <td>0.141559</td>\n",
       "      <td>21.087274</td>\n",
       "      <td>27.465195</td>\n",
       "      <td>0.389678</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>commercial</th>\n",
       "      <td>26.658957</td>\n",
       "      <td>33.085124</td>\n",
       "      <td>0.123129</td>\n",
       "      <td>19.882416</td>\n",
       "      <td>26.129721</td>\n",
       "      <td>0.251267</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>agricultural</th>\n",
       "      <td>27.750853</td>\n",
       "      <td>34.213460</td>\n",
       "      <td>0.148903</td>\n",
       "      <td>14.750685</td>\n",
       "      <td>19.559583</td>\n",
       "      <td>0.177105</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                 France                           Spain  \\\n",
       "                                    MAE       RMSE        R2        MAE   \n",
       "metatype        type                                                      \n",
       "residential     TH            18.092241  24.672725  0.431736  21.284345   \n",
       "                SFH           16.048118  21.978529  0.376297  16.545053   \n",
       "                MFH           20.716695  27.467044  0.309220  16.002098   \n",
       "                AB            15.622602  20.550609  0.035450  10.933419   \n",
       "non-residential others        21.866204  28.357705  0.276258        NaN   \n",
       "                industrial    15.761084  22.346697  0.141559  21.087274   \n",
       "                commercial    26.658957  33.085124  0.123129  19.882416   \n",
       "                agricultural  27.750853  34.213460  0.148903  14.750685   \n",
       "\n",
       "                                                   \n",
       "                                   RMSE        R2  \n",
       "metatype        type                               \n",
       "residential     TH            28.612939  0.275659  \n",
       "                SFH           22.494812  0.257353  \n",
       "                MFH           22.288014  0.296023  \n",
       "                AB            14.723366  0.396498  \n",
       "non-residential others              NaN       NaN  \n",
       "                industrial    27.465195  0.389678  \n",
       "                commercial    26.129721  0.251267  \n",
       "                agricultural  19.559583  0.177105  "
      ]
     },
     "execution_count": 287,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_country = df[df['NAME_0'].isin(['France', 'Spain'])]\n",
    "df_country = df_country.rename(columns={'NAME_0': 'Country'})\n",
    "\n",
    "df_country['metatype'] = df_country['type']\n",
    "# df_country['type'].fillna('unknown', inplace=True)\n",
    "df_country = df_country[df_country['type'] != 'unknown']\n",
    "df_country = harmonize_group_source_types(df_country)\n",
    "df_country['type'] = np.where(df_country['type'] == 'residential', df_country['residential_type'], df_country['type'])\n",
    "\n",
    "df_country_group = df_country.groupby(['Country', 'metatype', 'type'], dropna=True).apply(lambda x: pd.Series({\n",
    "        'MAE'       : metrics.mean_absolute_error(x['y_test'], x['y_predict']),\n",
    "        'RMSE'       : np.sqrt(metrics.mean_squared_error(x['y_test'], x['y_predict'])),\n",
    "        'R2'        : metrics.r2_score(x['y_test'], x['y_predict']),\n",
    "        'n'         : len(x.index),\n",
    "    })).reset_index()\n",
    "\n",
    "pivot_table = df_country_group.pivot_table(values=['R2', 'MAE', 'RMSE'], index=['metatype', 'type'], columns=['Country'], margins=False)\n",
    "pivot_table = pivot_table.swaplevel(0, 1, axis=1)\n",
    "pivot_table = pivot_table.sort_index(axis=1, level=0)\n",
    "pivot_table.columns = pivot_table.columns.rename(None, level=0) # drop Country cell\n",
    "\n",
    "new_cols = pivot_table.columns.reindex(['MAE', 'RMSE', 'R2'], level=1)\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "# pivot_table.loc[('France', 'All'),:] = df_country[df_country['Country'] == 'France'].groupby(['type'])['error'].apply(lambda x: x.abs().mean())\n",
    "# pivot_table.loc[('Spain', 'All'),:] = df_country[df_country['Country'] == 'Spain'].groupby(['type'])['error'].apply(lambda x: x.abs().mean())\n",
    "pivot_table = pivot_table.sort_index(ascending=False)\n",
    "\n",
    "pivot_table.to_latex('../figures/prediction-error-fine-granular-type.tex', index=True, float_format='%.2f', caption='Prediction error by fine-granular building type in France and Spain', label='table:prediction-error-fine-granular-types')\n",
    "pivot_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 222,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.145435"
      ]
     },
     "execution_count": 222,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "0.143941 + 0.000255 +0.000023 + 0.001216"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 324,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "NAME_0       type           \n",
       "France       residential        0.820773\n",
       "             non-residential    0.176885\n",
       "             unknown            0.002342\n",
       "Netherlands  unknown            1.000000\n",
       "Spain        residential        0.781580\n",
       "             non-residential    0.211977\n",
       "             unknown            0.006443\n",
       "Name: type, dtype: float64"
      ]
     },
     "execution_count": 324,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test = df\n",
    "df_test['type'].fillna('unknown', inplace=True)\n",
    "df_test['type_source'].fillna('unknown', inplace=True)\n",
    "# df_test.groupby('NAME_0')['type_source'].value_counts(normalize=True)\n",
    "df_test.groupby('NAME_0')['type'].value_counts(normalize=True)\n",
    "# df_test.groupby('NAME_0')['residential_type'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib qt\n",
    "df_test = df#.sample(frac=0.1)\n",
    "df_test = df_test[df_test['y_test'] <= 2020]\n",
    "df_test = df_test[df_test['y_predict'] <= 2030]\n",
    "# # df_test['DEGURBA'].replace({'1 (city)': 1, '2 (town)':2 , '3 (rural)': 3}, inplace=True)\n",
    "\n",
    "# # df_test['error'] = df_test['y_predict'] - df_test['y_test']\n",
    "df_test = df_test.rename(columns={'NAME_0': 'Country'})\n",
    "df_test['urban'] = df_test['DEGURBA'] < 3.0\n",
    "df_test.loc[df_test['type'] == 'non-residential', 'residential_type'] = 'non-residential'\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    sns.violinplot(data=df_test[df_test['Country'] == 'France'], x='residential_type', y='error', inner='quartile', split=True, palette='husl', order=['SFH', 'TH', 'MFH', 'AB', 'non-residential'], ax=ax)\n",
    "    ax.set_xlabel('Building type')\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "    plt.savefig('../figures/error-by-residential-type-FRA-violin.svg')\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    df_test['DEGURBA'].replace({1.0: '1 (city)', 2.0: '2 (town)', 3.0: '3 (rural)'}, inplace=True)\n",
    "    sns.violinplot(data=df_test[df_test['Country'] == 'France'], x='DEGURBA', y='error', hue='type', inner='quartile', split=True, palette='husl', order=['1 (city)', '2 (town)', '3 (rural)'], ax=ax)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "    plt.savefig('../figures/error-by-settlement-type-FRA-violin.svg')\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    sns.violinplot(data=df_test, x='Country', y='error', hue='urban', inner='quartile', split=True, palette='husl', order=['Netherlands', 'France', 'Spain'], ax=ax)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    df_test['DEGURBA'].replace({1.0: '1 (city)', 2.0: '2 (town)', 3.0: '3 (rural)'}, inplace=True)\n",
    "    sns.violinplot(data=df_test, x='DEGURBA', y='error', hue='Country', inner='quartile', split=False, order=['1 (city)', '2 (town)', '3 (rural)'], palette='husl', ax=ax)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "    ax.legend(title=None)\n",
    "    plt.title('Prediction error by settlement type')\n",
    "    plt.savefig('../figures/error-by-settlement-type-violin.svg')\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    sns.violinplot(data=df_test[df_test['type'].isin(['residential', 'non-residential'])], x='Country', y='error', hue='type', inner='quartile', split=True, palette='husl', order=['Netherlands', 'France', 'Spain'], ax=ax)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    df_test['type'].fillna('unknown', inplace=True)\n",
    "    order = ['residential', 'non-residential', 'unknown']\n",
    "    sns.violinplot(data=df_test[df_test['Country'].isin(['France', 'Spain'])], x='type', y='error', hue='Country', order=order, inner='quartile', split=True, palette='husl', ax=ax)\n",
    "    ax.set_xlabel('Building type', labelpad=20)\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "    ax.axvline(0.5, color='grey', linestyle='--', lw=0.75)\n",
    "    ax.axvline(1.5, color='grey', linestyle='--', lw=0.75)\n",
    "    plt.title('Prediction error for building types in France and Spain')\n",
    "    plt.savefig('../figures/error-by-type-country-violin.svg')\n",
    "\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager(figsize=(19, 9)) as ax: # 16,9 for Spain\n",
    "    order = ['SFH', 'TH', 'MFH', 'AB', 'commercial', 'agricultural', 'industrial', 'others', 'unknown']\n",
    "\n",
    "    country, annot_pos = 'France', [0.22, 0.67]\n",
    "    # country, annot_pos = 'Spain', [0.25, 0.69]\n",
    "    # order.remove('others')\n",
    "\n",
    "    df_country = df_test[df_test['Country'] == country]\n",
    "\n",
    "    df_country = harmonize_group_source_types(df_country)\n",
    "    df_country['type'].fillna('unknown', inplace=True)\n",
    "    df_country['type'] = np.where(df_country['type'] == 'residential', df_country['residential_type'], df_country['type'])\n",
    "\n",
    "    sns.violinplot(data=df_country, x='type', y='error', order=order, hue='Country', inner='quartile', split=True, palette='huel', ax=ax)\n",
    "\n",
    "    ax.annotate(text='residential\\nbuilding types', xy=(annot_pos[0],-0.26), xycoords='axes fraction', horizontalalignment='center', fontsize=20)\n",
    "    ax.annotate(text='non-residential\\nbuilding types', xy=(annot_pos[1],-0.26), xycoords='axes fraction', horizontalalignment='center', fontsize=20)\n",
    "    ax.axvline(3.5, color='grey', linestyle='--', lw=0.75)\n",
    "    ax.axvline(len(order) - 1.5, color='grey', linestyle='--', lw=0.75)\n",
    "\n",
    "    ax.set_ylim([-120, 120])\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "    ax.set_xlabel(None)\n",
    "    ax.set_xticklabels(['single-family house', 'terraced house', 'multi-family house', 'apartment block', 'commercial', 'agricultural', 'industrial', 'others', 'unknown'])\n",
    "    plt.xticks(rotation=30)\n",
    "    plt.title(f'Prediction error for fine-granular building types in {country}')\n",
    "    plt.savefig(f'../figures/error-by-type-{country}-violin.svg')\n",
    "\n",
    "    df_country['type'].value_counts()\n",
    "\n",
    "\n",
    "with visualizations.SubplotManager(figsize=(19, 9)) as ax: # 16,9 for Spain\n",
    "    order = ['SFH', 'TH', 'MFH', 'AB', 'commercial', 'agricultural', 'industrial', 'others']\n",
    "    labels = ['single-family house', 'terraced house', 'multi-family house', 'apartment block', 'commercial', 'agricultural', 'industrial', 'others']\n",
    "    colors = ['thistle', 'lightseagreen']\n",
    "\n",
    "    df_country = df_test[df_test['Country'].isin(['Spain', 'France'])]\n",
    "\n",
    "    df_country = harmonize_group_source_types(df_country)\n",
    "    df_country = df_country[df_country['type'] != 'unknown']\n",
    "    df_country['type'] = np.where(df_country['type'] == 'residential', df_country['residential_type'], df_country['type'])\n",
    "\n",
    "    sns.violinplot(data=df_country, x='type', y='error', order=order, hue='Country', inner='quartile', split=True, palette=colors, ax=ax)\n",
    "\n",
    "    ax.annotate(text='residential\\nbuilding types', xy=(0.25,-0.35), xycoords='axes fraction', horizontalalignment='center', fontsize=20)\n",
    "    ax.annotate(text='non-residential\\nbuilding types', xy=(0.70,-0.35), xycoords='axes fraction', horizontalalignment='center', fontsize=20)\n",
    "    ax.axvline(3.5, color='grey', linestyle='--', lw=0.75)\n",
    "\n",
    "    ax.set_ylim([-120, 120])\n",
    "    ax.set_ylabel('Prediction error [years]')\n",
    "    ax.set_xlabel(None)\n",
    "    shares = df_country.groupby('Country')['type'].value_counts(normalize=True).mul(100).round(1).sort_index(level=0).groupby('type').apply(list).to_dict()\n",
    "    shares['others'].append(0)\n",
    "    labels = [f'{labels[i]}\\n({shares[subtype][0]}% | {shares[subtype][1]}%)' for i, subtype in enumerate(order)]\n",
    "    ax.set_xticklabels(labels)\n",
    "    # ax.set_aspect(aspect=0.021)\n",
    "    plt.xticks(rotation=30)\n",
    "    plt.title(f'Prediction error for fine-granular building types in France and Spain', pad=20)\n",
    "    plt.savefig(f'../figures/error-by-finegranular-type-violin.svg')\n",
    "\n",
    "    df_country['type'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 383,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "NAME_0  type           \n",
       "France  residential        0.822699\n",
       "        non-residential    0.177301\n",
       "Spain   residential        0.786648\n",
       "        non-residential    0.213352\n",
       "Name: type, dtype: float64"
      ]
     },
     "execution_count": 383,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df['type'] != 'unknown'].groupby('NAME_0')['type'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 373,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'AB': [0.2, 1.8],\n",
       " 'MFH': [1.9, 3.8],\n",
       " 'SFH': [38.1, 23.2],\n",
       " 'TH': [42.2, 49.6],\n",
       " 'agricultural': [0.1, 8.1],\n",
       " 'commercial': [2.7, 3.8],\n",
       " 'industrial': [0.5, 9.7],\n",
       " 'others': [14.4]}"
      ]
     },
     "execution_count": 373,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_country.groupby('Country')['type'].value_counts(normalize=True).mul(100).round(1).sort_index(level=0).groupby('type').apply(list).to_dict() #.swaplevel(0, 1).reset_index(level=1, drop=True).to_dict() #.mul(100).round(1).to_dict()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 249,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Résidentiel               1467894\n",
       "Annexe                     257429\n",
       "Commercial et services      48272\n",
       "Industriel                   7974\n",
       "Indifférencié                4188\n",
       "Agricole                     2175\n",
       "Sportif                       456\n",
       "Religieux                      41\n",
       "Name: type_source, dtype: int64"
      ]
     },
     "execution_count": 249,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df['NAME_0'] == 'France']['type_source'].value_counts(dropna=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 250,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1_residential         1199850\n",
       "3_industrial           146895\n",
       "2_agriculture          121998\n",
       "4_3_publicServices      29639\n",
       "4_2_retail              21281\n",
       "unknown                  9891\n",
       "4_1_office               5606\n",
       "Name: type_source, dtype: int64"
      ]
     },
     "execution_count": 250,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df['NAME_0'] == 'Spain']['type_source'].value_counts(dropna=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 323,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "residential           18520\n",
       "industrial             2378\n",
       "agricultural            579\n",
       "commercial              442\n",
       "fine-granular-type       16\n",
       "Name: type, dtype: int64"
      ]
     },
     "execution_count": 323,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "harm_df = harmonize_group_source_types(df_es_sample)\n",
    "harm_df[['type', 'type_source']]['type'].value_counts(dropna=False)\n",
    "# harm_df[harm_df[dataset.TYPE_ATTRIBUTE] != 'residential'][dataset.TYPE_ATTRIBUTE].isin(['residential', None])\n",
    "# harm_df[dataset.TYPE_ATTRIBUTE].isin([None]).sort_values()\n",
    "harm_df[dataset.TYPE_ATTRIBUTE] = np.where((harm_df[dataset.TYPE_ATTRIBUTE].isnull()), 'fine-granular-type', harm_df[dataset.TYPE_ATTRIBUTE])\n",
    "harm_df[['type', 'type_source']]['type'].value_counts(dropna=False)\n",
    "\n",
    "# categorical_to_int(harm_df, var='type', labels=['residential', 'commercial', 'agricultural', 'industrial'])[['type', 'type_source', 'type_label']]['type'].value_counts()\n",
    "# df_es_sample['type'].unique()\n",
    "# df_es_sample['type'].astype('category').cat.codes\n",
    "# categorical_to_int(remove_buildings_with_unknown_type(df_fr_sample), var='type', labels=['residential', 'non-residential'])[['type', 'type_source', 'type_label', 'residential_type']]['type'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### City comparison"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 464,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>MAE mean</th>\n",
       "      <th>MAE std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <th>DEGURBA</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">France</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>124</td>\n",
       "      <td>18.357471</td>\n",
       "      <td>4.931862</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>315</td>\n",
       "      <td>16.374822</td>\n",
       "      <td>3.766594</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>2083</td>\n",
       "      <td>20.267423</td>\n",
       "      <td>3.975355</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>2725</td>\n",
       "      <td>19.718968</td>\n",
       "      <td>4.294601</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">Netherlands</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>98</td>\n",
       "      <td>14.910172</td>\n",
       "      <td>4.449907</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>238</td>\n",
       "      <td>15.282632</td>\n",
       "      <td>4.000469</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>88</td>\n",
       "      <td>16.590624</td>\n",
       "      <td>2.411283</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>424</td>\n",
       "      <td>15.468015</td>\n",
       "      <td>3.880323</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">Spain</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>107</td>\n",
       "      <td>15.523778</td>\n",
       "      <td>4.551334</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>640</td>\n",
       "      <td>17.399037</td>\n",
       "      <td>5.121387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>3530</td>\n",
       "      <td>26.591560</td>\n",
       "      <td>9.136878</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>5260</td>\n",
       "      <td>24.923218</td>\n",
       "      <td>9.257727</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                       count   MAE mean   MAE std\n",
       "Country     DEGURBA                              \n",
       "France      1 (city)     124  18.357471  4.931862\n",
       "            2 (town)     315  16.374822  3.766594\n",
       "            3 (rural)   2083  20.267423  3.975355\n",
       "            All         2725  19.718968  4.294601\n",
       "Netherlands 1 (city)      98  14.910172  4.449907\n",
       "            2 (town)     238  15.282632  4.000469\n",
       "            3 (rural)     88  16.590624  2.411283\n",
       "            All          424  15.468015  3.880323\n",
       "Spain       1 (city)     107  15.523778  4.551334\n",
       "            2 (town)     640  17.399037  5.121387\n",
       "            3 (rural)   3530  26.591560  9.136878\n",
       "            All         5260  24.923218  9.257727"
      ]
     },
     "execution_count": 464,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "\n",
    "error_dfs = []\n",
    "# bins = utils.generate_bins((1900, 2020, 10))\n",
    "bins = dataset.TABULA_AGE_BINS['harmonized']\n",
    "labels = utils.generate_labels(bins)\n",
    "for predictor, city_boundaries, gadm_level, dissolve_level in [(regressor_fts_importance_fr, gadm_france, 4, 2), (regressor_fts_importance_es, gadm_spain, 4, 2), (regressor_fts_importance_nl, gadm_netherlands, 2, 1)]:\n",
    "    df = predictor.individual_prediction_error()\n",
    "    df['error'] = df['error'].abs()\n",
    "    df['y_test'] = predictor.y_test[dataset.AGE_ATTRIBUTE]\n",
    "    df['y_predict'] = predictor.y_predict[dataset.AGE_ATTRIBUTE]\n",
    "    df['period'] = pd.cut(predictor.y_test[dataset.AGE_ATTRIBUTE], bins, right=False).cat.codes\n",
    "    df['period'].replace(dict(zip(range(1,8), labels[1:])), inplace=True)\n",
    "    print(f'Mean error: {df[\"error\"].mean()}')\n",
    "    df = df.join(predictor.aux_vars_test, how=\"inner\")\n",
    "\n",
    "    error_by_city = df.groupby(['city'], dropna=False).apply(lambda x: pd.Series({\n",
    "            'MAE'       : x['error'].mean(),\n",
    "            'MAE-metrics': metrics.mean_absolute_error(x['y_test'], x['y_predict']),\n",
    "            'n'         : len(x.index),\n",
    "            'R2'        : metrics.r2_score(x['y_test'], x['y_predict']) # doesn't make sense for grouping on period\n",
    "        })).reset_index()\n",
    "\n",
    "    error_by_city = error_by_city.merge(city_boundaries[[f'NAME_{gadm_level}', 'NAME_0', 'DEGURBA']], right_on=f'NAME_{gadm_level}', left_on='city', how='left')\n",
    "    error_dfs.append(error_by_city)\n",
    "\n",
    "error_df = pd.concat(error_dfs)\n",
    "\n",
    "# column renaming\n",
    "error_df['DEGURBA'].replace({1.0: '1 (city)', 2.0: '2 (town)', 3.0: '3 (rural)'}, inplace=True)\n",
    "error_df = error_df.rename(columns={'NAME_0': 'Country'})\n",
    "\n",
    "table = error_df.groupby(['Country', 'DEGURBA'])['MAE'].agg(['count', 'mean', 'std'])\n",
    "table_total = error_df.groupby(['Country'])['MAE'].agg(['count', 'mean', 'std'])\n",
    "table_total['DEGURBA'] = 'All'\n",
    "new_table = pd.concat([table, table_total.set_index(['DEGURBA'], append=True)]).sort_index(level=0)\n",
    "new_table = new_table.rename(columns={'mean': 'MAE mean', 'std': 'MAE std', 'count': 'n'})\n",
    "new_table.to_latex('../figures/prediction-error-by-city-type.tex', index=True, float_format='%.2f', caption='Prediction error by settlement type', label='table:error-by-settlement-type')\n",
    "\n",
    "new_table\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 472,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>France</th>\n",
       "      <td>2725</td>\n",
       "      <td>19.718968</td>\n",
       "      <td>4.294601</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Netherlands</th>\n",
       "      <td>424</td>\n",
       "      <td>15.468015</td>\n",
       "      <td>3.880323</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Spain</th>\n",
       "      <td>5260</td>\n",
       "      <td>24.923218</td>\n",
       "      <td>9.257727</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             count       mean       std\n",
       "Country                                \n",
       "France        2725  19.718968  4.294601\n",
       "Netherlands    424  15.468015  3.880323\n",
       "Spain         5260  24.923218  9.257727"
      ]
     },
     "execution_count": 472,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "error_df.groupby(['Country'])['MAE'].agg(['count', 'mean', 'std'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 490,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <th>DEGURBA</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">France</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>113</td>\n",
       "      <td>18.337403</td>\n",
       "      <td>4.715487</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>293</td>\n",
       "      <td>16.383979</td>\n",
       "      <td>3.583438</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>2049</td>\n",
       "      <td>20.222279</td>\n",
       "      <td>3.849010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">Netherlands</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>96</td>\n",
       "      <td>14.810744</td>\n",
       "      <td>4.425439</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>232</td>\n",
       "      <td>15.004623</td>\n",
       "      <td>3.407316</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>88</td>\n",
       "      <td>16.590624</td>\n",
       "      <td>2.411283</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">Spain</th>\n",
       "      <th>1 (city)</th>\n",
       "      <td>97</td>\n",
       "      <td>15.526334</td>\n",
       "      <td>4.645267</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2 (town)</th>\n",
       "      <td>594</td>\n",
       "      <td>17.320194</td>\n",
       "      <td>4.776974</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3 (rural)</th>\n",
       "      <td>2681</td>\n",
       "      <td>26.095696</td>\n",
       "      <td>7.777249</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                       count       mean       std\n",
       "Country     DEGURBA                              \n",
       "France      1 (city)     113  18.337403  4.715487\n",
       "            2 (town)     293  16.383979  3.583438\n",
       "            3 (rural)   2049  20.222279  3.849010\n",
       "Netherlands 1 (city)      96  14.810744  4.425439\n",
       "            2 (town)     232  15.004623  3.407316\n",
       "            3 (rural)     88  16.590624  2.411283\n",
       "Spain       1 (city)      97  15.526334  4.645267\n",
       "            2 (town)     594  17.320194  4.776974\n",
       "            3 (rural)   2681  26.095696  7.777249"
      ]
     },
     "execution_count": 490,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "error_df[error_df['n'] > 100].groupby(['Country', 'DEGURBA'])['MAE'].agg(['count', 'mean', 'std'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 486,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "< 10         0.029658\n",
       "< 100        0.211217\n",
       "< 1000       0.697909\n",
       "< 10000      0.980228\n",
       "< 100000     0.999810\n",
       "< 1000000    1.000000\n",
       "Name: n, dtype: float64"
      ]
     },
     "execution_count": 486,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.cumsum(pd.cut(error_df[error_df['Country'] == 'Spain']['n'], [0, 10, 100, 1000, 10000, 100000, 1000000], labels=['< 10','< 100', '< 1000', '< 10000', '< 100000','< 1000000'], right=False).value_counts(normalize=True).sort_index())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 512,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(-0.22173416920097944, 4.03027800035525e-06)\n",
      "(-0.37208220329527203, 3.1655677891026756e-90)\n",
      "(-0.2188654878862992, 4.5045282385866765e-58)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Please use `pearsonr` from the `scipy.stats` namespace, the `scipy.stats.stats` namespace is deprecated.\n"
     ]
    }
   ],
   "source": [
    "from scipy import stats\n",
    "for c in ['Netherlands', 'France', 'Spain']:\n",
    "    df = error_df[(error_df['Country'] == c)]\n",
    "    print(stats.stats.pearsonr(df['MAE'], df['n']))\n",
    "    stats.linregress(df['MAE'], df['n'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "'NAME_0'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/indexes/base.py:3621\u001b[0m, in \u001b[0;36mIndex.get_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   3620\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m-> 3621\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_engine\u001b[39m.\u001b[39;49mget_loc(casted_key)\n\u001b[1;32m   3622\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mKeyError\u001b[39;00m \u001b[39mas\u001b[39;00m err:\n",
      "File \u001b[0;32mpandas/_libs/index.pyx:136\u001b[0m, in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/index.pyx:163\u001b[0m, in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/hashtable_class_helper.pxi:5198\u001b[0m, in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/hashtable_class_helper.pxi:5206\u001b[0m, in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mKeyError\u001b[0m: 'NAME_0'",
      "\nThe above exception was the direct cause of the following exception:\n",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb Cell 116'\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000172?line=0'>1</a>\u001b[0m \u001b[39mimport\u001b[39;00m \u001b[39mseaborn\u001b[39;00m \u001b[39mas\u001b[39;00m \u001b[39msns\u001b[39;00m\n\u001b[1;32m      <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000172?line=1'>2</a>\u001b[0m \u001b[39m# error_df[error_df['R2'] < 0] = 0\u001b[39;00m\n\u001b[0;32m----> <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000172?line=2'>3</a>\u001b[0m df_nl_scatter \u001b[39m=\u001b[39m error_df[(error_df[\u001b[39m'\u001b[39m\u001b[39mR2\u001b[39m\u001b[39m'\u001b[39m] \u001b[39m>\u001b[39m \u001b[39m0\u001b[39m) \u001b[39m&\u001b[39m (error_df[\u001b[39m'\u001b[39;49m\u001b[39mNAME_0\u001b[39;49m\u001b[39m'\u001b[39;49m] \u001b[39m==\u001b[39m \u001b[39m'\u001b[39m\u001b[39mNetherlands\u001b[39m\u001b[39m'\u001b[39m)]\n\u001b[1;32m      <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000172?line=3'>4</a>\u001b[0m ax \u001b[39m=\u001b[39m sns\u001b[39m.\u001b[39mlmplot(x\u001b[39m=\u001b[39m\u001b[39m'\u001b[39m\u001b[39mn\u001b[39m\u001b[39m'\u001b[39m, y\u001b[39m=\u001b[39m\u001b[39m'\u001b[39m\u001b[39mR2\u001b[39m\u001b[39m'\u001b[39m, data\u001b[39m=\u001b[39mdf_nl_scatter, hue\u001b[39m=\u001b[39m\u001b[39m'\u001b[39m\u001b[39mDEGURBA\u001b[39m\u001b[39m'\u001b[39m, scatter\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m, fit_reg\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m, scatter_kws\u001b[39m=\u001b[39m{\u001b[39m\"\u001b[39m\u001b[39ms\u001b[39m\u001b[39m\"\u001b[39m: \u001b[39m1\u001b[39m})\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/frame.py:3505\u001b[0m, in \u001b[0;36mDataFrame.__getitem__\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   3503\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mcolumns\u001b[39m.\u001b[39mnlevels \u001b[39m>\u001b[39m \u001b[39m1\u001b[39m:\n\u001b[1;32m   3504\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_getitem_multilevel(key)\n\u001b[0;32m-> 3505\u001b[0m indexer \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mcolumns\u001b[39m.\u001b[39;49mget_loc(key)\n\u001b[1;32m   3506\u001b[0m \u001b[39mif\u001b[39;00m is_integer(indexer):\n\u001b[1;32m   3507\u001b[0m     indexer \u001b[39m=\u001b[39m [indexer]\n",
      "File \u001b[0;32m/usr/local/Caskroom/miniconda/base/envs/uf-ml/lib/python3.8/site-packages/pandas/core/indexes/base.py:3623\u001b[0m, in \u001b[0;36mIndex.get_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   3621\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_engine\u001b[39m.\u001b[39mget_loc(casted_key)\n\u001b[1;32m   3622\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mKeyError\u001b[39;00m \u001b[39mas\u001b[39;00m err:\n\u001b[0;32m-> 3623\u001b[0m     \u001b[39mraise\u001b[39;00m \u001b[39mKeyError\u001b[39;00m(key) \u001b[39mfrom\u001b[39;00m \u001b[39merr\u001b[39;00m\n\u001b[1;32m   3624\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mTypeError\u001b[39;00m:\n\u001b[1;32m   3625\u001b[0m     \u001b[39m# If we have a listlike key, _check_indexing_error will raise\u001b[39;00m\n\u001b[1;32m   3626\u001b[0m     \u001b[39m#  InvalidIndexError. Otherwise we fall through and re-raise\u001b[39;00m\n\u001b[1;32m   3627\u001b[0m     \u001b[39m#  the TypeError.\u001b[39;00m\n\u001b[1;32m   3628\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_check_indexing_error(key)\n",
      "\u001b[0;31mKeyError\u001b[0m: 'NAME_0'"
     ]
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "# error_df[error_df['R2'] < 0] = 0\n",
    "df_nl_scatter = error_df[(error_df['R2'] > 0) & (error_df['NAME_0'] == 'Netherlands')]\n",
    "ax = sns.lmplot(x='n', y='R2', data=df_nl_scatter, hue='DEGURBA', scatter=True, fit_reg=True, scatter_kws={\"s\": 1})\n",
    "# ax.set(xscale=\"log\")\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Figure: Error on Map"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from sklearn import metrics\n",
    "# %matplotlib qt\n",
    "# label = False\n",
    "# # visualizations.plot_prediction_error_on_map_by_city(regressor_fts_importance_nl, geometry.to_gdf(gadm_netherlands), dissolve_level=None)\n",
    "# # visualizations.plot_prediction_error_on_map_by_city(regressor_fts_importance_fr, geometry.to_gdf(gadm_france), gadm_level=4, dissolve_level=None)\n",
    "# # visualizations.plot_prediction_error_on_map_by_city(regressor_fts_importance_es, geometry.to_gdf(gadm_spain), gadm_level=4, dissolve_level=2)\n",
    "# import matplotlib\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    error_dfs = []\n",
    "    for predictor, city_boundaries, gadm_level, dissolve_level in [(regressor_fts_importance_fr, gadm_france, 4, 2), (regressor_fts_importance_es, gadm_spain, 4, 2), (regressor_fts_importance_nl, gadm_netherlands, 2, 1)]:\n",
    "    # for predictor, city_boundaries, gadm_level, dissolve_level in [(regressor_fts_importance_es, gadm_spain, 4, 2)]:\n",
    "    # for predictor, city_boundaries, gadm_level, dissolve_level in [(regressor_fts_importance_nl, gadm_netherlands, 2, 1)]:\n",
    "        y = predictor.individual_prediction_error()\n",
    "        y['error'] = y['error'].abs()\n",
    "        y['y_test'] = predictor.y_test[dataset.AGE_ATTRIBUTE]\n",
    "        y['y_predict'] = predictor.y_predict[dataset.AGE_ATTRIBUTE]\n",
    "        y = pd.concat([y, predictor.aux_vars_test], axis=1, join=\"inner\")\n",
    "\n",
    "        y = y.merge(city_boundaries[[f'NAME_{gadm_level}', f'NAME_{dissolve_level}']], right_on=f'NAME_{gadm_level}', left_on='city', how='left').reset_index()\n",
    "\n",
    "        error_by_city = y.groupby([f'NAME_{dissolve_level}'], dropna=False).apply(lambda x: pd.Series({\n",
    "            # 'error'       : x['error'].mean(),\n",
    "            'mae': metrics.mean_absolute_error(x['y_test'], x['y_predict']),\n",
    "            'n'            : len(x.index),\n",
    "            'error'        : metrics.r2_score(x['y_test'], x['y_predict'])\n",
    "        })).reset_index()\n",
    "        # error_by_city = y.groupby('city')['error'].mean()\n",
    "        # error_by_city = y.groupby('city').apply(lambda x: metrics.r2_score(x['y_test'], x['y_predict'])).reset_index(name='error')\n",
    "        error_by_city[error_by_city['error'] < 0] = 0\n",
    "        error_by_city = geometry.to_gdf(city_boundaries).dissolve(by=f'GID_{dissolve_level}').merge(error_by_city, on=f'NAME_{dissolve_level}', how='left')\n",
    "        error_by_city['region'] = error_by_city[f'NAME_{dissolve_level}']\n",
    "\n",
    "        if label:\n",
    "            min = error_by_city.iloc[error_by_city['error'].idxmin()]\n",
    "            max = error_by_city.iloc[error_by_city['error'].idxmax()]\n",
    "            min_text = f\"min error {min['error']:.2f}\\n({min[f'NAME_{gadm_level}']})\"\n",
    "            max_text = f\"max error {max['error']:.2f}\\n({max[f'NAME_{gadm_level}']})\"\n",
    "            ax.annotate(text=min_text, xy=min.geometry.centroid.coords[0], ha='center')\n",
    "            ax.annotate(text=max_text, xy=max.geometry.centroid.coords[0], ha='center')\n",
    "\n",
    "        error_dfs.append(error_by_city)\n",
    "    error_df = pd.concat(error_dfs)\n",
    "\n",
    "    colors = ['plum', 'thistle', 'lavenderblush', 'paleturquoise', 'mediumturquoise']\n",
    "    # colors = ['plum', 'thistle', 'lavenderblush', 'paleturquoise', 'mediumturquoise', 'lightseagreen']\n",
    "    # colors = list(reversed(colors))\n",
    "    cmap = matplotlib.colors.ListedColormap(colors)\n",
    "\n",
    "    error_df.dissolve('GID_0').boundary.plot(ax=ax, color='lavender', linewidth=0.3)\n",
    "    error_df.plot(ax=ax, column='error', legend=True, cmap=cmap, edgecolor=None, linewidth=0, legend_kwds={'shrink': 0.7, 'orientation': 'vertical', 'pad': 0,  'label': '$R^2$'})\n",
    "    plt.axis('off')\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 987,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>geometry</th>\n",
       "      <th>NAME_0</th>\n",
       "      <th>NAME_1</th>\n",
       "      <th>NAME_2</th>\n",
       "      <th>NAME_3</th>\n",
       "      <th>NAME_4</th>\n",
       "      <th>GID_0</th>\n",
       "      <th>GID_1</th>\n",
       "      <th>GID_3</th>\n",
       "      <th>GID_4</th>\n",
       "      <th>...</th>\n",
       "      <th>LAU CODE</th>\n",
       "      <th>POPULATION</th>\n",
       "      <th>NL_NAME_1</th>\n",
       "      <th>GID_2</th>\n",
       "      <th>VARNAME_2</th>\n",
       "      <th>NL_NAME_2</th>\n",
       "      <th>TYPE_2</th>\n",
       "      <th>ENGTYPE_2</th>\n",
       "      <th>CC_2</th>\n",
       "      <th>HASC_2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>POLYGON ((3769576.539 2492100.687, 3768086.046...</td>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Puy-de-Dôme</td>\n",
       "      <td>Ambert</td>\n",
       "      <td>Ambert</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.10.1_1</td>\n",
       "      <td>FRA.1.10.1.1_1</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>POLYGON ((3909714.070 2499142.030, 3909428.161...</td>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Rhône</td>\n",
       "      <td>Lyon</td>\n",
       "      <td>Bron</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.11.1_1</td>\n",
       "      <td>FRA.1.11.1.1_1</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>POLYGON ((4021093.098 2479134.719, 4020595.965...</td>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Savoie</td>\n",
       "      <td>Albertville</td>\n",
       "      <td>Aime</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.12.1_1</td>\n",
       "      <td>FRA.1.12.1.1_1</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>POLYGON ((3930615.784 2534620.365, 3929853.199...</td>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Ain</td>\n",
       "      <td>Belley</td>\n",
       "      <td>Ambérieu-en-Bugey</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.1.1_1</td>\n",
       "      <td>FRA.1.1.1.1_1</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>POLYGON ((3788893.861 2576368.672, 3788516.717...</td>\n",
       "      <td>France</td>\n",
       "      <td>Auvergne-Rhône-Alpes</td>\n",
       "      <td>Allier</td>\n",
       "      <td>Montluçon</td>\n",
       "      <td>Commentry</td>\n",
       "      <td>FRA</td>\n",
       "      <td>FRA.1_1</td>\n",
       "      <td>FRA.1.2.1_1</td>\n",
       "      <td>FRA.1.2.1.2_1</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>MULTIPOLYGON (((4099138.277 3339437.378, 40986...</td>\n",
       "      <td>Netherlands</td>\n",
       "      <td>Groningen</td>\n",
       "      <td>Appingedam</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>...</td>\n",
       "      <td>GM0003</td>\n",
       "      <td>11721.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD.5.1_1</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Gemeente</td>\n",
       "      <td>Municipality</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NL.GR.AP</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>POLYGON ((3980611.667 3260569.929, 3980229.804...</td>\n",
       "      <td>Netherlands</td>\n",
       "      <td>IJsselmeer</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>...</td>\n",
       "      <td>GM0532</td>\n",
       "      <td>21706.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD.6.1_1</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Water body</td>\n",
       "      <td>Water body</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>POLYGON ((4020112.859 3081146.867, 4020081.923...</td>\n",
       "      <td>Netherlands</td>\n",
       "      <td>Limburg</td>\n",
       "      <td>Beek</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>...</td>\n",
       "      <td>GM0888</td>\n",
       "      <td>15929.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD.7.3_1</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Gemeente</td>\n",
       "      <td>Municipality</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NL.LI.BK</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>POLYGON ((3922602.953 3157673.632, 3922783.633...</td>\n",
       "      <td>Netherlands</td>\n",
       "      <td>Noord-Brabant</td>\n",
       "      <td>'s-Hertogenbosch</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>...</td>\n",
       "      <td>GM0796</td>\n",
       "      <td>154205.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD.8.1_1</td>\n",
       "      <td>Bergeyk</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Gemeente</td>\n",
       "      <td>Municipality</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NL.NB.SH</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>MULTIPOLYGON (((3958430.234 3248441.037, 39582...</td>\n",
       "      <td>Netherlands</td>\n",
       "      <td>Noord-Holland</td>\n",
       "      <td>Aalsmeer</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>...</td>\n",
       "      <td>GM0358</td>\n",
       "      <td>31728.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NLD.9.1_1</td>\n",
       "      <td>Haarlemmerliede en Spaarnwoude</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Gemeente</td>\n",
       "      <td>Municipality</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NL.NH.AA</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>162 rows × 35 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             geometry       NAME_0  \\\n",
       "0   POLYGON ((3769576.539 2492100.687, 3768086.046...       France   \n",
       "1   POLYGON ((3909714.070 2499142.030, 3909428.161...       France   \n",
       "2   POLYGON ((4021093.098 2479134.719, 4020595.965...       France   \n",
       "3   POLYGON ((3930615.784 2534620.365, 3929853.199...       France   \n",
       "4   POLYGON ((3788893.861 2576368.672, 3788516.717...       France   \n",
       "..                                                ...          ...   \n",
       "9   MULTIPOLYGON (((4099138.277 3339437.378, 40986...  Netherlands   \n",
       "10  POLYGON ((3980611.667 3260569.929, 3980229.804...  Netherlands   \n",
       "11  POLYGON ((4020112.859 3081146.867, 4020081.923...  Netherlands   \n",
       "12  POLYGON ((3922602.953 3157673.632, 3922783.633...  Netherlands   \n",
       "13  MULTIPOLYGON (((3958430.234 3248441.037, 39582...  Netherlands   \n",
       "\n",
       "                  NAME_1            NAME_2       NAME_3             NAME_4  \\\n",
       "0   Auvergne-Rhône-Alpes       Puy-de-Dôme       Ambert             Ambert   \n",
       "1   Auvergne-Rhône-Alpes             Rhône         Lyon               Bron   \n",
       "2   Auvergne-Rhône-Alpes            Savoie  Albertville               Aime   \n",
       "3   Auvergne-Rhône-Alpes               Ain       Belley  Ambérieu-en-Bugey   \n",
       "4   Auvergne-Rhône-Alpes            Allier    Montluçon          Commentry   \n",
       "..                   ...               ...          ...                ...   \n",
       "9              Groningen        Appingedam          NaN                NaN   \n",
       "10            IJsselmeer              None          NaN                NaN   \n",
       "11               Limburg              Beek          NaN                NaN   \n",
       "12         Noord-Brabant  's-Hertogenbosch          NaN                NaN   \n",
       "13         Noord-Holland          Aalsmeer          NaN                NaN   \n",
       "\n",
       "   GID_0    GID_1         GID_3           GID_4  ... LAU CODE POPULATION  \\\n",
       "0    FRA  FRA.1_1  FRA.1.10.1_1  FRA.1.10.1.1_1  ...      NaN        NaN   \n",
       "1    FRA  FRA.1_1  FRA.1.11.1_1  FRA.1.11.1.1_1  ...      NaN        NaN   \n",
       "2    FRA  FRA.1_1  FRA.1.12.1_1  FRA.1.12.1.1_1  ...      NaN        NaN   \n",
       "3    FRA  FRA.1_1   FRA.1.1.1_1   FRA.1.1.1.1_1  ...      NaN        NaN   \n",
       "4    FRA  FRA.1_1   FRA.1.2.1_1   FRA.1.2.1.2_1  ...      NaN        NaN   \n",
       "..   ...      ...           ...             ...  ...      ...        ...   \n",
       "9    NLD      NaN           NaN             NaN  ...   GM0003    11721.0   \n",
       "10   NLD      NaN           NaN             NaN  ...   GM0532    21706.0   \n",
       "11   NLD      NaN           NaN             NaN  ...   GM0888    15929.0   \n",
       "12   NLD      NaN           NaN             NaN  ...   GM0796   154205.0   \n",
       "13   NLD      NaN           NaN             NaN  ...   GM0358    31728.0   \n",
       "\n",
       "   NL_NAME_1      GID_2                       VARNAME_2  NL_NAME_2  \\\n",
       "0        NaN        NaN                             NaN        NaN   \n",
       "1        NaN        NaN                             NaN        NaN   \n",
       "2        NaN        NaN                             NaN        NaN   \n",
       "3        NaN        NaN                             NaN        NaN   \n",
       "4        NaN        NaN                             NaN        NaN   \n",
       "..       ...        ...                             ...        ...   \n",
       "9        NaN  NLD.5.1_1                            None        NaN   \n",
       "10       NaN  NLD.6.1_1                            None        NaN   \n",
       "11       NaN  NLD.7.3_1                            None        NaN   \n",
       "12       NaN  NLD.8.1_1                         Bergeyk        NaN   \n",
       "13       NaN  NLD.9.1_1  Haarlemmerliede en Spaarnwoude        NaN   \n",
       "\n",
       "        TYPE_2     ENGTYPE_2  CC_2    HASC_2  \n",
       "0          NaN           NaN   NaN       NaN  \n",
       "1          NaN           NaN   NaN       NaN  \n",
       "2          NaN           NaN   NaN       NaN  \n",
       "3          NaN           NaN   NaN       NaN  \n",
       "4          NaN           NaN   NaN       NaN  \n",
       "..         ...           ...   ...       ...  \n",
       "9     Gemeente  Municipality   NaN  NL.GR.AP  \n",
       "10  Water body    Water body   NaN      None  \n",
       "11    Gemeente  Municipality   NaN  NL.LI.BK  \n",
       "12    Gemeente  Municipality   NaN  NL.NB.SH  \n",
       "13    Gemeente  Municipality   NaN  NL.NH.AA  \n",
       "\n",
       "[162 rows x 35 columns]"
      ]
     },
     "execution_count": 987,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# error_df[['NAME_0', 'NAME_', 'mae', 'error']]\n",
    "error_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 635,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>region</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ESP_A Coruña_80</td>\n",
       "      <td>0.338157</td>\n",
       "      <td>17.797119</td>\n",
       "      <td>A Coruña</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ESP_Albacete_80</td>\n",
       "      <td>0.398620</td>\n",
       "      <td>23.863497</td>\n",
       "      <td>Albacete</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>ESP_Alicante_80</td>\n",
       "      <td>0.562238</td>\n",
       "      <td>12.096494</td>\n",
       "      <td>Alicante</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>ESP_Almería_80</td>\n",
       "      <td>0.536436</td>\n",
       "      <td>13.291172</td>\n",
       "      <td>Almería</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ESP_Asturias_80</td>\n",
       "      <td>0.389359</td>\n",
       "      <td>21.869520</td>\n",
       "      <td>Asturias</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ESP_Badajoz_80</td>\n",
       "      <td>0.379609</td>\n",
       "      <td>19.369742</td>\n",
       "      <td>Badajoz</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>ESP_Baleares_80</td>\n",
       "      <td>0.510720</td>\n",
       "      <td>15.231747</td>\n",
       "      <td>Baleares</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>ESP_Barcelona_80</td>\n",
       "      <td>0.494997</td>\n",
       "      <td>15.157180</td>\n",
       "      <td>Barcelona</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ESP_Burgos_80</td>\n",
       "      <td>0.414159</td>\n",
       "      <td>22.632680</td>\n",
       "      <td>Burgos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>ESP_Cantabria_80</td>\n",
       "      <td>0.451652</td>\n",
       "      <td>20.498453</td>\n",
       "      <td>Cantabria</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>ESP_Castellón_80</td>\n",
       "      <td>0.530828</td>\n",
       "      <td>15.997301</td>\n",
       "      <td>Castellón</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                name        R2        MAE     region\n",
       "0    ESP_A Coruña_80  0.338157  17.797119   A Coruña\n",
       "3    ESP_Albacete_80  0.398620  23.863497   Albacete\n",
       "10   ESP_Alicante_80  0.562238  12.096494   Alicante\n",
       "9     ESP_Almería_80  0.536436  13.291172    Almería\n",
       "2    ESP_Asturias_80  0.389359  21.869520   Asturias\n",
       "1     ESP_Badajoz_80  0.379609  19.369742    Badajoz\n",
       "7    ESP_Baleares_80  0.510720  15.231747   Baleares\n",
       "6   ESP_Barcelona_80  0.494997  15.157180  Barcelona\n",
       "4      ESP_Burgos_80  0.414159  22.632680     Burgos\n",
       "5   ESP_Cantabria_80  0.451652  20.498453  Cantabria\n",
       "8   ESP_Castellón_80  0.530828  15.997301  Castellón"
      ]
     },
     "execution_count": 635,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "spe = pd.read_csv('../notebooks/specialized-regional-models-23915991-20220730-01-01-42-itermediate-comparison-results.csv')[['name', 'R2', 'MAE']]\n",
    "# add_features_results['Country'] = add_features_results['name'].str.split('_').str[0]\n",
    "spe['region'] = spe['name'].str.split('_').str[1]\n",
    "spe.sort_values(by='region', ascending=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 639,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>NAME_1</th>\n",
       "      <th>error</th>\n",
       "      <th>mae</th>\n",
       "      <th>n</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Drenthe</td>\n",
       "      <td>0.365694</td>\n",
       "      <td>14.231995</td>\n",
       "      <td>237026.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Flevoland</td>\n",
       "      <td>0.548705</td>\n",
       "      <td>6.525441</td>\n",
       "      <td>174893.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Friesland</td>\n",
       "      <td>0.451444</td>\n",
       "      <td>14.988049</td>\n",
       "      <td>263894.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>Gelderland</td>\n",
       "      <td>0.367535</td>\n",
       "      <td>15.285617</td>\n",
       "      <td>773655.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>Groningen</td>\n",
       "      <td>0.451543</td>\n",
       "      <td>16.364632</td>\n",
       "      <td>267191.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       NAME_1     error        mae         n\n",
       "5     Drenthe  0.365694  14.231995  237026.0\n",
       "6   Flevoland  0.548705   6.525441  174893.0\n",
       "7   Friesland  0.451444  14.988049  263894.0\n",
       "8  Gelderland  0.367535  15.285617  773655.0\n",
       "9   Groningen  0.451543  16.364632  267191.0"
      ]
     },
     "execution_count": 639,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# error_df[error_df['NAME_2'].isin(spe['region'].values)][['NAME_2', 'error', 'mae', 'n']]\n",
    "error_df[error_df['NAME_1'].isin(['Flevoland', 'Groningen', 'Friesland', 'Gelderland', 'Drenthe'])][['NAME_1', 'error', 'mae', 'n']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 272,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.07050298567984417"
      ]
     },
     "execution_count": 272,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "error_df[error_df['NAME_0'] == 'France'].sort_values(by='error')[['error', 'NAME_2']].dropna()['error'].std()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "availability\n",
      "80    9.617869\n",
      "dtype: float64\n",
      "1.0\n",
      "availability\n",
      "80    16.659844\n",
      "dtype: float64\n",
      "0.6808510638297872\n",
      "availability\n",
      "80    15.79765\n",
      "dtype: float64\n",
      "0.967032967032967\n"
     ]
    }
   ],
   "source": [
    "for results in [regional_nld, regional_esp, regional_fra]:\n",
    "    results['availability'] = results['name'].str.split('_').str[0].astype(int)\n",
    "    results['Country'] = results['name'].str.split('_').str[1]\n",
    "    results['Region'] = results['name'].str.split('_').str[2]\n",
    "    results['n'] = results['name'].str.split('_').str[3].astype(int)\n",
    "    results = results[results['availability'] == 80]\n",
    "    print(results.groupby('availability').apply(lambda x: np.average(x['MAE'], weights=x.n)))\n",
    "    print(sum(results['MAE'] < 20) / len(results))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3845554.2428076672, 4149464.4738966064, 3059900.564596048, 3404339.508031441)"
      ]
     },
     "execution_count": 163,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# plot with predefined value bins / value color mapping\n",
    "colors = ['plum', 'thistle', 'lavenderblush', 'paleturquoise', 'mediumturquoise']\n",
    "error_df['error_color'] = pd.cut(error_df['error'], [-10, 0, 0.25, 0.5, 0.75, 1], labels=['< 0.00', '0.00 - 0.25', '0.25 - 0.50', '0.50 - 0.75', '> 0.75'], right=False)\n",
    "cmap = matplotlib.colors.ListedColormap(list(reversed(colors)))\n",
    "error_df.plot(legend=True, column='error_color', cmap=cmap, edgecolor=None, linewidth=0)\n",
    "plt.axis('off')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>country</th>\n",
       "      <th>MAE</th>\n",
       "      <th>R2</th>\n",
       "      <th>&lt;10</th>\n",
       "      <th>&lt;15</th>\n",
       "      <th>&lt;20</th>\n",
       "      <th>n</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ESP</td>\n",
       "      <td>17.477905</td>\n",
       "      <td>0.428060</td>\n",
       "      <td>0.407354</td>\n",
       "      <td>0.558425</td>\n",
       "      <td>0.673693</td>\n",
       "      <td>7758976.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>FRA</td>\n",
       "      <td>16.841685</td>\n",
       "      <td>0.455593</td>\n",
       "      <td>0.449828</td>\n",
       "      <td>0.597259</td>\n",
       "      <td>0.700729</td>\n",
       "      <td>9596614.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>NLD</td>\n",
       "      <td>10.084359</td>\n",
       "      <td>0.632857</td>\n",
       "      <td>0.685579</td>\n",
       "      <td>0.778479</td>\n",
       "      <td>0.838263</td>\n",
       "      <td>5441430.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  country        MAE        R2       <10       <15       <20          n\n",
       "0     ESP  17.477905  0.428060  0.407354  0.558425  0.673693  7758976.0\n",
       "1     FRA  16.841685  0.455593  0.449828  0.597259  0.700729  9596614.0\n",
       "2     NLD  10.084359  0.632857  0.685579  0.778479  0.838263  5441430.0"
      ]
     },
     "execution_count": 131,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# from sklearn import metrics\n",
    "\n",
    "# # national_nld = AgePredictor.load('../results/model-NLD-random-cv.pkl')\n",
    "# # national_esp = AgePredictor.load('../results/model-ESP-random-cv.pkl')\n",
    "# # national_fra = AgePredictor.load('../results/model-FRA-random-cv.pkl')\n",
    "\n",
    "# results = []\n",
    "# for predictor in [national_nld, national_esp, national_fra]:\n",
    "#     country_result = predictor.aux_vars_test[['city', 'country']]\n",
    "#     country_result['y_test'] = predictor.y_test[dataset.AGE_ATTRIBUTE]\n",
    "#     country_result['y_predict'] = predictor.y_predict[dataset.AGE_ATTRIBUTE]\n",
    "#     # country_result = country_result.merge(gadm[[f'NAME_{gadm_level}', f'NAME_{dissolve_level}']], left_on='city', right_on=f'NAME_{gadm_level}', how='left').reset_index()\n",
    "#     results.append(country_result)\n",
    "\n",
    "# country_result = pd.concat(results, axis=0)\n",
    "\n",
    "# country_result_grouped = country_result.groupby(['country'], dropna=False).apply(lambda x: pd.Series({\n",
    "#         'MAE'       : metrics.mean_absolute_error(x['y_test'], x['y_predict']),\n",
    "#         'R2'        : metrics.r2_score(x['y_test'], x['y_predict']),\n",
    "#         '<10'        : sum((x['y_test'] - x['y_predict']).abs() < 10) / len(x),\n",
    "#         '<15'        : sum((x['y_test'] - x['y_predict']).abs() < 15) / len(x),\n",
    "#         '<20'        : sum((x['y_test'] - x['y_predict']).abs() < 20) / len(x),\n",
    "#         'n'            : len(x.index),\n",
    "#     })).reset_index()\n",
    "country_result_grouped"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Specialized regional models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "ename": "",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31mThe Kernel crashed while executing code in the the current cell or a previous cell. Please review the code in the cell(s) to identify a possible cause of the failure. Click <a href='https://aka.ms/vscodeJupyterKernelCrash'>here</a> for more info. View Jupyter <a href='command:jupyter.viewOutput'>log</a> for further details."
     ]
    }
   ],
   "source": [
    "# from sklearn import metrics\n",
    "\n",
    "# regional_nld = pd.read_csv('../results/results-specialized-regional-models-NLD.csv')\n",
    "# regional_esp = pd.read_csv('../results/results-specialized-regional-models-ESP.csv')\n",
    "# regional_fra = pd.read_csv('../results/results-specialized-regional-models-FRA.csv')\n",
    "\n",
    "# # national_nld = AgePredictor.load('../results/model-NLD-random-cv.pkl')\n",
    "# # national_esp = AgePredictor.load('../results/model-ESP-random-cv.pkl')\n",
    "# # national_fra = AgePredictor.load('../results/model-FRA-random-cv.pkl')\n",
    "\n",
    "\n",
    "\n",
    "# import matplotlib\n",
    "# geo_results = []\n",
    "# for predictor, results, gadm, gadm_level, dissolve_level in [(national_fra, regional_fra, gadm_france, 4, 2), (national_esp, regional_esp, gadm_spain, 4, 2), (national_nld, regional_nld, gadm_netherlands, 2, 1)]:\n",
    "#     results['Local data availability'] = results['name'].str.split('_').str[0].astype(int)\n",
    "#     results['Country'] = results['name'].str.split('_').str[1]\n",
    "#     results['Region'] = results['name'].str.split('_').str[2]\n",
    "#     results['n'] = results['name'].str.split('_').str[3].astype(int)\n",
    "\n",
    "#     country_result = predictor.aux_vars_test[['city', 'country']]\n",
    "#     country_result['y_test'] = predictor.y_test[dataset.AGE_ATTRIBUTE]\n",
    "#     country_result['y_predict'] = predictor.y_predict[dataset.AGE_ATTRIBUTE]\n",
    "#     country_result = country_result.merge(gadm[[f'NAME_{gadm_level}', f'NAME_{dissolve_level}']], left_on='city', right_on=f'NAME_{gadm_level}', how='left').reset_index()\n",
    "\n",
    "#     country_result_by_region = country_result.groupby(['country', f'NAME_{dissolve_level}'], dropna=False).apply(lambda x: pd.Series({\n",
    "#             'MAE'       : metrics.mean_absolute_error(x['y_test'], x['y_predict']),\n",
    "#             'R2'        : metrics.r2_score(x['y_test'], x['y_predict']),\n",
    "#             'n'            : len(x.index),\n",
    "#         })).reset_index()\n",
    "#     country_result_by_region['Region'] = country_result_by_region[f'NAME_{dissolve_level}']\n",
    "#     country_result_by_region['Country'] = country_result_by_region['country']\n",
    "#     country_result_by_region['Local data availability'] = 'non-regional'\n",
    "\n",
    "#     gadm = geometry.to_gdf(gadm).dissolve(by=f'GID_{dissolve_level}')\n",
    "#     results = pd.concat([results, country_result_by_region], axis=0)\n",
    "#     geo_results.append(gadm.merge(results, left_on=f'NAME_{dissolve_level}', right_on='Region', how='left').reset_index())\n",
    "\n",
    "# eu_geo_results = pd.concat(geo_results, axis=0)\n",
    "# borders = eu_geo_results.dissolve('GID_0').boundary.simplify(1000)\n",
    "\n",
    "# # colors = ['plum', 'thistle', 'lavenderblush', 'paleturquoise', 'mediumturquoise']\n",
    "# colors = ['plum', 'thistle', 'lavenderblush', 'paleturquoise', 'mediumturquoise', 'lightseagreen']\n",
    "# # # colors = list(reversed(colors))\n",
    "# cmap = matplotlib.colors.ListedColormap(colors)\n",
    "\n",
    "\n",
    "##### BOXPLOT #####\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    # ax.set_ylim(0, 1)\n",
    "    # sns.lineplot(data=eu_geo_results.reset_index(drop=True).sort_values(by=['Local data availability','R2']), hue='Local data availability', y='R2', x='Region', ax=ax)\n",
    "    # sns.boxplot(data=eu_geo_results.reset_index(drop=True).sort_values(by=['Local data availability','R2']), y='Local data availability', x='R2', hue='Country', palette=['darksalmon', 'steelblue', 'teal'], ax=ax)\n",
    "\n",
    "    # best visualization (besides map)\n",
    "    ax.set_xlim(0, 1)\n",
    "    prepared_eu_geo_results = eu_geo_results.replace({'non-regional': 'national model\\n(random cv)', 10: '10%', 20: '20%', 50: '50%', 80: '80%'}).reset_index(drop=True).sort_values(by=['Local data availability','R2'])\n",
    "    sns.boxplot(data=prepared_eu_geo_results, fliersize=0, hue='Local data availability', x='R2', y='Country', palette='husl', hue_order=['10%', '20%', '50%', '80%', 'national model\\n(random cv)'], ax=ax)\n",
    "    ax.set_yticklabels(['Spain', 'France', 'Netherlands'])\n",
    "    ax.set_ylabel(None)\n",
    "    ax.set_xlabel('$R^2$')\n",
    "    plt.savefig('../figures/regional-models-boxplot.svg')\n",
    "\n",
    "\n",
    "##### TABLE #####\n",
    "def f1(x):\n",
    "    return '%.1f' % x\n",
    "\n",
    "def f2(x):\n",
    "    return '%.2f' % x\n",
    "\n",
    "# table = prepared_eu_geo_results.groupby(['Country', 'Local data availability'])[['R2', 'MAE']].mean().unstack(level=0).swaplevel(0,1, axis=1).sort_index(axis=1)\n",
    "# table.to_latex('../figures/regional-models.tex', index=True,  formatters=[f1, f2, f1, f2, f1, f2], caption='Impact of local data availability for region models', label='table:regional-models')\n",
    "\n",
    "\n",
    "##### MAP #####\n",
    "# with visualizations.SubplotManager(ncols=3, constrained_layout=True) as ax:\n",
    "#     for idx, share in enumerate([10, 80, 'non-regional']):\n",
    "#         results_x = eu_geo_results[eu_geo_results['Local data availability'] == share]\n",
    "#         # results_x = results_x[results_x['Country'] == 'FRA']\n",
    "\n",
    "#         # results_x['R2_pos'] = utils.custom_round(results_x['R2'].apply(lambda x: x if x > 0 else 0), base=0.2)\n",
    "#         # results_x['R2_pos'] = pd.cut(results_x['R2'], bins=[-10,0.2,0.4,0.6,0.8])\n",
    "\n",
    "\n",
    "#         # cmap_boundaries = matplotlib.colors.ListedColormap(['darksalmon', 'steelblue', 'teal'])\n",
    "#         # gpd.GeoDataFrame(geometry=eu_geo_results.dissolve('GID_0').boundary.simplify(1000)).reset_index().plot(ax=ax[idx], column='GID_0', cmap=cmap_boundaries, linewidth=0.3)\n",
    "\n",
    "#         borders.plot(ax=ax[idx], color='lightgray', linewidth=0.3)\n",
    "#         results_x.plot(ax=ax[idx], column='R2', legend=idx==2, vmin=0.09, vmax=0.61, cmap=cmap, edgecolor=None, linewidth=0, legend_kwds={'shrink': 0.4, 'orientation': 'vertical', 'pad': 0.15, 'label': 'R2', 'ticks': [0.1, 0.2, 0.3, 0.4, 0.5]})\n",
    "#         # results_x.plot(ax=ax, column='MAE', legend=True, vmin=5, vmax=15, cmap=cmap, edgecolor=None, linewidth=0)\n",
    "#         ax[idx].set_axis_off()\n",
    "#         ax[idx].set_title(f'regional model\\n({share}% local data)' if share != 'non-regional' else 'national model\\n(80% data)', fontsize=14)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1256,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Country     \n",
       "ESP      MAE   -1.028366\n",
       "         R2     0.032135\n",
       "FRA      MAE   -1.375989\n",
       "         R2     0.040198\n",
       "NLD      MAE   -0.686609\n",
       "         R2     0.023715\n",
       "dtype: float64"
      ]
     },
     "execution_count": 1256,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "table.loc['80%'] - table.loc['national model\\n(random cv)']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "with visualizations.SubplotManager() as ax:\n",
    "\n",
    "    geometry.to_gdf(gadm_france).dissolve('GID_2').plot(ax=ax)\n",
    "    geometry.to_gdf(gadm_france).dissolve('GID_2').apply(lambda x: ax.annotate(text=x['NAME_2'], xy=x.geometry.centroid.coords[0], ha='center'), axis=1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 944,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>n</th>\n",
       "      <th>Country</th>\n",
       "      <th>Local data availability</th>\n",
       "      <th>R2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>167099</td>\n",
       "      <td>ESP</td>\n",
       "      <td>10</td>\n",
       "      <td>0.446093</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>167099</td>\n",
       "      <td>ESP</td>\n",
       "      <td>20</td>\n",
       "      <td>0.477795</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>167099</td>\n",
       "      <td>ESP</td>\n",
       "      <td>50</td>\n",
       "      <td>0.513747</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>167099</td>\n",
       "      <td>ESP</td>\n",
       "      <td>80</td>\n",
       "      <td>0.530615</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>287993</td>\n",
       "      <td>ESP</td>\n",
       "      <td>10</td>\n",
       "      <td>0.470725</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>1035722</td>\n",
       "      <td>NLD</td>\n",
       "      <td>80</td>\n",
       "      <td>0.577262</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>570857</td>\n",
       "      <td>NLD</td>\n",
       "      <td>10</td>\n",
       "      <td>0.622760</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>570857</td>\n",
       "      <td>NLD</td>\n",
       "      <td>20</td>\n",
       "      <td>0.654139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48</th>\n",
       "      <td>570857</td>\n",
       "      <td>NLD</td>\n",
       "      <td>50</td>\n",
       "      <td>0.686171</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49</th>\n",
       "      <td>570857</td>\n",
       "      <td>NLD</td>\n",
       "      <td>80</td>\n",
       "      <td>0.699561</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>243 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "          n Country Local data availability        R2\n",
       "0    167099     ESP                      10  0.446093\n",
       "1    167099     ESP                      20  0.477795\n",
       "2    167099     ESP                      50  0.513747\n",
       "3    167099     ESP                      80  0.530615\n",
       "4    287993     ESP                      10  0.470725\n",
       "..      ...     ...                     ...       ...\n",
       "45  1035722     NLD                      80  0.577262\n",
       "46   570857     NLD                      10  0.622760\n",
       "47   570857     NLD                      20  0.654139\n",
       "48   570857     NLD                      50  0.686171\n",
       "49   570857     NLD                      80  0.699561\n",
       "\n",
       "[243 rows x 4 columns]"
      ]
     },
     "execution_count": 944,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eu_geo_results[['n', 'Country', 'Local data availability', 'R2']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### In-fill housing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 409,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Passing a set as an indexer is deprecated and will raise in a future version. Use a list instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Netherlands\n",
      "25.544858895949314\n",
      "14.250209838073424\n",
      "-1.6008990583185478\n",
      "0.44387218709115106\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Passing a set as an indexer is deprecated and will raise in a future version. Use a list instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Spain\n",
      "36.290889297123165\n",
      "19.15466827433844\n",
      "-0.4462284106901184\n",
      "0.3606706108260671\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Passing a set as an indexer is deprecated and will raise in a future version. Use a list instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "France\n",
      "43.68887679819899\n",
      "17.934925873480832\n",
      "-8.164556310258924\n",
      "0.41130697697658825\n"
     ]
    }
   ],
   "source": [
    "import ast\n",
    "predictors = {'Netherlands': regressor_fts_importance_nl, 'Spain': regressor_fts_importance_es, 'France': regressor_fts_importance_fr}\n",
    "for country, predictor in predictors.items():\n",
    "    y_test = predictor.y_test.rename(columns={'age': 'y_test'})\n",
    "    y_predict = predictor.y_predict.rename(columns={'age': 'y_predict'})\n",
    "    y = pd.concat([y_test, y_predict, predictor.aux_vars_test], axis=1, join=\"inner\")\n",
    "    y = determine_infill_housing(y)\n",
    "    print(country)\n",
    "    print(metrics.mean_absolute_error(y[y['infill']]['y_test'], y[y['infill']]['y_predict']))\n",
    "    print(metrics.mean_absolute_error(y[~y['infill']]['y_test'], y[~y['infill']]['y_predict']))\n",
    "    print(metrics.r2_score(y[y['infill']]['y_test'], y[y['infill']]['y_predict']))\n",
    "    print(metrics.r2_score(y[~y['infill']]['y_test'], y[~y['infill']]['y_predict']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 410,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "18.063333333333336"
      ]
     },
     "execution_count": 410,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(25.54 - 14.25 + 36.29 - 19.15 + 43.69 - 17.93) / 3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 417,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3.9499999999999993"
      ]
     },
     "execution_count": 417,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(16.59 - 14.31 + 20.27 - 18.00 + 26.59 - 19.29) / 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 251,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Passing a set as an indexer is deprecated and will raise in a future version. Use a list instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-17 17:56:42,844 | WARNING : Dataset is already a GeoDataFrame. Skipping convertion.\n"
     ]
    }
   ],
   "source": [
    "def determine_infill_housing(df):\n",
    "    columns = list(df.columns)\n",
    "    df['n_neighbors'] = df['TouchesIndexes'].fillna('[1]').apply(lambda l: len(ast.literal_eval(l)))\n",
    "    groups = df.groupby('sbb')['y_test']\n",
    "    groups_mean = groups.transform('mean')\n",
    "    groups_std = groups.transform('std')\n",
    "    groups_size = groups.transform('size')\n",
    "    df['sbb_mean'] = groups_mean\n",
    "    df['sbb_deviation'] = df['y_test'] - groups_mean\n",
    "    df['sbb_std'] = groups_std\n",
    "    df['outliers'] = ((df['sbb_deviation'] / groups_std).abs() > 2) & (groups_size > 10)\n",
    "    df['infill'] = (df['sbb_deviation'] / groups_std > 2) & (groups_size > 10) & (df['n_neighbors'] > 3)\n",
    "    return df[set(columns + ['infill'])]\n",
    "\n",
    "predictor = regressor_fts_importance_fr\n",
    "y = predictor.y_test.rename(columns={'age': 'y_test'})\n",
    "y = pd.concat([y, predictor.aux_vars_test], axis=1, join=\"inner\")\n",
    "y = y[y['city'] == 'Vichy']\n",
    "y_geom = determine_infill_housing(y)\n",
    "y_geom['id'] = y_geom['id_source']\n",
    "y_geom = y_geom.drop(columns=['geometry'])\n",
    "y_geom = geometry.add_geometry_column(y_geom, crs=3035, countries=['france'])\n",
    "y_geom = geometry.to_gdf(y_geom)\n",
    "\n",
    "with visualizations.SubplotManager() as ax:\n",
    "    y_geom.plot(ax=ax, column='y_test', legend=True, cmap='Purples', edgecolor=None)\n",
    "    # y_geom[y_geom['outliers']].plot(ax=ax, color='red')\n",
    "    y_geom[y_geom['infill']].boundary.plot(ax=ax, color='orange')\n",
    "    plt.axis('off')\n",
    "\n",
    "\n",
    "# y[['y_test', 'y_predict', 'sbb_deviation', 'infill', 'sbb_mean', 'sbb_std', 'sbb']]\n",
    "# y['infill'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature importance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Table: Group feature importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Netherlands: 36 of 120 features are significant for regression.\n",
      "France: 29 of 120 features are significant for regression.\n",
      "Spain: 24 of 120 features are significant for regression.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Netherlands</th>\n",
       "      <th>France</th>\n",
       "      <th>Spain</th>\n",
       "      <th>All</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Building</th>\n",
       "      <td>0.13</td>\n",
       "      <td>0.23</td>\n",
       "      <td>0.28</td>\n",
       "      <td>0.213333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Building neighborhood</th>\n",
       "      <td>0.26</td>\n",
       "      <td>0.19</td>\n",
       "      <td>0.19</td>\n",
       "      <td>0.213333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block</th>\n",
       "      <td>0.12</td>\n",
       "      <td>0.09</td>\n",
       "      <td>0.14</td>\n",
       "      <td>0.116667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block neighborhood</th>\n",
       "      <td>0.13</td>\n",
       "      <td>0.08</td>\n",
       "      <td>0.09</td>\n",
       "      <td>0.100000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Street</th>\n",
       "      <td>0.06</td>\n",
       "      <td>0.13</td>\n",
       "      <td>0.07</td>\n",
       "      <td>0.086667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Street neighborhood</th>\n",
       "      <td>0.05</td>\n",
       "      <td>0.07</td>\n",
       "      <td>0.05</td>\n",
       "      <td>0.056667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Street-based block</th>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.010000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Street-based block neighborhood</th>\n",
       "      <td>0.05</td>\n",
       "      <td>0.04</td>\n",
       "      <td>0.03</td>\n",
       "      <td>0.040000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Street centrality</th>\n",
       "      <td>0.09</td>\n",
       "      <td>0.10</td>\n",
       "      <td>0.05</td>\n",
       "      <td>0.080000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City</th>\n",
       "      <td>0.09</td>\n",
       "      <td>0.06</td>\n",
       "      <td>0.09</td>\n",
       "      <td>0.080000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total: all building</th>\n",
       "      <td>0.39</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.47</td>\n",
       "      <td>0.426667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total: all block</th>\n",
       "      <td>0.25</td>\n",
       "      <td>0.17</td>\n",
       "      <td>0.23</td>\n",
       "      <td>0.216667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total: all street</th>\n",
       "      <td>0.20</td>\n",
       "      <td>0.30</td>\n",
       "      <td>0.17</td>\n",
       "      <td>0.223333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total: all street-based blocks</th>\n",
       "      <td>0.06</td>\n",
       "      <td>0.05</td>\n",
       "      <td>0.04</td>\n",
       "      <td>0.050000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total: building &amp; block &amp; street</th>\n",
       "      <td>0.33</td>\n",
       "      <td>0.46</td>\n",
       "      <td>0.51</td>\n",
       "      <td>0.433333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total: neighborhood &amp; city</th>\n",
       "      <td>0.67</td>\n",
       "      <td>0.54</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.566667</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                   Netherlands  France  Spain       All\n",
       "Building                                  0.13    0.23   0.28  0.213333\n",
       "Building neighborhood                     0.26    0.19   0.19  0.213333\n",
       "Block                                     0.12    0.09   0.14  0.116667\n",
       "Block neighborhood                        0.13    0.08   0.09  0.100000\n",
       "Street                                    0.06    0.13   0.07  0.086667\n",
       "Street neighborhood                       0.05    0.07   0.05  0.056667\n",
       "Street-based block                        0.01    0.01   0.01  0.010000\n",
       "Street-based block neighborhood           0.05    0.04   0.03  0.040000\n",
       "Street centrality                         0.09    0.10   0.05  0.080000\n",
       "City                                      0.09    0.06   0.09  0.080000\n",
       "Total: all building                       0.39    0.42   0.47  0.426667\n",
       "Total: all block                          0.25    0.17   0.23  0.216667\n",
       "Total: all street                         0.20    0.30   0.17  0.223333\n",
       "Total: all street-based blocks            0.06    0.05   0.04  0.050000\n",
       "Total: building & block & street          0.33    0.46   0.51  0.433333\n",
       "Total: neighborhood & city                0.67    0.54   0.49  0.566667"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%matplotlib qt\n",
    "predictors = {'Netherlands': regressor_fts_importance_nl, 'France': regressor_fts_importance_fr, 'Spain': regressor_fts_importance_es}\n",
    "# predictors = {'France': regressor_fts_importance_fr, 'Netherlands': regressor_fts_importance_nl, 'Spain': regressor_fts_importance_es}\n",
    "\n",
    "df_importance = []\n",
    "df_group_importance = []\n",
    "for country, predictor in predictors.items():\n",
    "    fts_importance_reg = predictor.normalized_feature_importance()\n",
    "    df_importance.append(fts_importance_reg)\n",
    "\n",
    "    fts_importance_reg_sig = fts_importance_reg[fts_importance_reg['normalized_importance'] > 0.01]\n",
    "    print(f'{country}: {len(fts_importance_reg_sig)} of {len(fts_importance_reg)} features are significant for regression.')\n",
    "    # neighborhood_mask = (fts_importance_reg['feature'].str.contains('within_buffer')) | (fts_importance_reg['feature'].isin(dataset.CITY_FEATURES))\n",
    "    # street_mask = fts_importance_reg['feature'].isin(dataset.STREET_FEATURES)\n",
    "    # building_mask = fts_importance_reg['feature'].isin(dataset.BUILDING_FEATURES + dataset.BLOCK_FEATURES)\n",
    "    # neighborhood_fts_importance = round(sum(fts_importance_reg.loc[neighborhood_mask].normalized_importance), 2)\n",
    "    # street_fts_importance = round(sum(fts_importance_reg.loc[street_mask].normalized_importance), 2)\n",
    "    # building_fts_importance = round(sum(fts_importance_reg.loc[building_mask].normalized_importance), 2)\n",
    "    # print(f'{country}: Share of feature contribution captured by neighborhood features {neighborhood_fts_importance}')\n",
    "    # print(f'{country}: Share of feature contribution captured by building features {building_fts_importance}')\n",
    "    # print(f'{country}: Share of feature contribution captured by street features {street_fts_importance}')\n",
    "    fts_groups = {\n",
    "        'Building': dataset.BUILDING_FEATURES,\n",
    "        'Building neighborhood': dataset.BUILDING_FEATURES_NEIGHBORHOOD,\n",
    "        'Block': dataset.BLOCK_FEATURES,\n",
    "        'Block neighborhood': dataset.BLOCK_FEATURES_NEIGHBORHOOD,\n",
    "        # 'Street': dataset.STREET_FEATURES,\n",
    "        # 'Street neighborhood': dataset.STREET_FEATURES_NEIGHBORHOOD,\n",
    "        'Street': dataset.STREET_ONLY_FEATURES,\n",
    "        'Street neighborhood': dataset.STREET_ONLY_FEATURES_NEIGHBORHOOD,\n",
    "        'Street-based block': dataset.SBB_FEATURES,\n",
    "        'Street-based block neighborhood': dataset.SBB_FEATURES_NEIGHBORHOOD,\n",
    "        'Street centrality': dataset.STREET_FEATURES_CENTRALITY,\n",
    "        'City': dataset.CITY_FEATURES,\n",
    "        'Total: all building': dataset.BUILDING_FEATURES_ALL,\n",
    "        'Total: all block': dataset.BLOCK_FEATURES_ALL,\n",
    "        # 'Total: all street ': dataset.STREET_FEATURES_ALL,\n",
    "        'Total: all street ': dataset.STREET_ONLY_FEATURES + dataset.STREET_ONLY_FEATURES_NEIGHBORHOOD + dataset.STREET_FEATURES_CENTRALITY,\n",
    "        'Total: all street-based blocks ': dataset.SBB_FEATURES + dataset.SBB_FEATURES_NEIGHBORHOOD,\n",
    "        # 'Total: building & block & street ': dataset.BUILDING_FEATURES + dataset.BLOCK_FEATURES + dataset.STREET_FEATURES,\n",
    "        'Total: building & block & street ': dataset.BUILDING_FEATURES + dataset.BLOCK_FEATURES + dataset.STREET_ONLY_FEATURES + dataset.SBB_FEATURES,\n",
    "        'Total: neighborhood & city':\n",
    "            dataset.BUILDING_FEATURES_NEIGHBORHOOD\n",
    "            + dataset.BLOCK_FEATURES_NEIGHBORHOOD\n",
    "            + dataset.STREET_ONLY_FEATURES_NEIGHBORHOOD\n",
    "            + dataset.SBB_FEATURES_NEIGHBORHOOD\n",
    "            + dataset.STREET_FEATURES_CENTRALITY\n",
    "            + dataset.CITY_FEATURES,\n",
    "        # 'Total': dataset.BUILDING_FEATURES +\n",
    "        #         dataset.BUILDING_FEATURES_NEIGHBORHOOD +\n",
    "        #         dataset.BLOCK_FEATURES +\n",
    "        #         dataset.BLOCK_FEATURES_NEIGHBORHOOD +\n",
    "        #         dataset.STREET_FEATURES +\n",
    "        #         dataset.STREET_FEATURES_NEIGHBORHOOD +\n",
    "        #         dataset.STREET_FEATURES_CENTRALITY +\n",
    "        #         dataset.CITY_FEATURES,\n",
    "    }\n",
    "    importance = {}\n",
    "    for name, features in fts_groups.items():\n",
    "        importance[name] = round(sum(fts_importance_reg.loc[fts_importance_reg[\"feature\"].isin(features)].normalized_importance), 2)\n",
    "\n",
    "    df_group_importance.append(pd.Series(importance, name=country))\n",
    "\n",
    "df = pd.concat(df_importance, axis=0, keys=list(predictors.keys())).reset_index(level=[0])\n",
    "df_group = pd.concat(df_group_importance, axis=1)\n",
    "\n",
    "# top_10_features = list(df.groupby('feature').mean().sort_values('normalized_importance', ascending=False).head(10).index)\n",
    "# combined_top_5_features = set().union(*[df[:5]['feature'].values for df in df_importance])\n",
    "# visualizations.slope_chart(df_importance, labels=list(predictors.keys()), feature_selection=top_10_features)\n",
    "\n",
    "df_group['All'] = df_group.mean(axis=1)\n",
    "df_group.to_latex('../figures/feature-importance-groups.tex', index=True, float_format='%.2f', caption='Normalized feature importance by feature group', label='table:fts-importance-groups')\n",
    "df_group\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Table: Individual feature importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>Country</th>\n",
       "      <th>Netherlands</th>\n",
       "      <th>France</th>\n",
       "      <th>Spain</th>\n",
       "      <th>All</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>feature</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>distance_to_closest_street</th>\n",
       "      <td>0.032212</td>\n",
       "      <td>0.078079</td>\n",
       "      <td>0.022058</td>\n",
       "      <td>0.044116</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>lat</th>\n",
       "      <td>0.022805</td>\n",
       "      <td>0.043192</td>\n",
       "      <td>0.062895</td>\n",
       "      <td>0.042964</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>FootprintArea</th>\n",
       "      <td>0.011029</td>\n",
       "      <td>0.057100</td>\n",
       "      <td>0.050449</td>\n",
       "      <td>0.039526</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>StdBlockFootprintArea</th>\n",
       "      <td>0.026739</td>\n",
       "      <td>0.013229</td>\n",
       "      <td>0.077640</td>\n",
       "      <td>0.039203</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>buildings_within_buffer_100</th>\n",
       "      <td>0.025147</td>\n",
       "      <td>0.037694</td>\n",
       "      <td>0.042311</td>\n",
       "      <td>0.035051</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>street_betweeness_global_closest_street</th>\n",
       "      <td>0.031033</td>\n",
       "      <td>0.050493</td>\n",
       "      <td>0.018063</td>\n",
       "      <td>0.033196</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SharedWallLength</th>\n",
       "      <td>0.017817</td>\n",
       "      <td>0.011597</td>\n",
       "      <td>0.058996</td>\n",
       "      <td>0.029470</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>total_ft_area_within_buffer_500</th>\n",
       "      <td>0.028431</td>\n",
       "      <td>0.016130</td>\n",
       "      <td>0.028131</td>\n",
       "      <td>0.024231</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>lon</th>\n",
       "      <td>0.010297</td>\n",
       "      <td>0.017277</td>\n",
       "      <td>0.036905</td>\n",
       "      <td>0.021493</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std_elongation_within_buffer_100</th>\n",
       "      <td>0.041104</td>\n",
       "      <td>0.004552</td>\n",
       "      <td>0.014099</td>\n",
       "      <td>0.019918</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>street_width_av_closest_street</th>\n",
       "      <td>0.007405</td>\n",
       "      <td>0.022217</td>\n",
       "      <td>0.027337</td>\n",
       "      <td>0.018986</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>av_footprint_area_within_buffer_100</th>\n",
       "      <td>0.015932</td>\n",
       "      <td>0.031816</td>\n",
       "      <td>0.007144</td>\n",
       "      <td>0.018297</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Phi</th>\n",
       "      <td>0.017194</td>\n",
       "      <td>0.032414</td>\n",
       "      <td>0.005182</td>\n",
       "      <td>0.018263</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Convexity</th>\n",
       "      <td>0.020747</td>\n",
       "      <td>0.007709</td>\n",
       "      <td>0.022656</td>\n",
       "      <td>0.017037</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>total_ft_area_within_buffer_100</th>\n",
       "      <td>0.017750</td>\n",
       "      <td>0.010914</td>\n",
       "      <td>0.019386</td>\n",
       "      <td>0.016017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std_convexity_within_buffer_100</th>\n",
       "      <td>0.014546</td>\n",
       "      <td>0.013240</td>\n",
       "      <td>0.016621</td>\n",
       "      <td>0.014802</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>av_convexity_within_buffer_100</th>\n",
       "      <td>0.011077</td>\n",
       "      <td>0.021118</td>\n",
       "      <td>0.011193</td>\n",
       "      <td>0.014463</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BlockConvexity</th>\n",
       "      <td>0.012685</td>\n",
       "      <td>0.021373</td>\n",
       "      <td>0.008645</td>\n",
       "      <td>0.014234</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BlockTotalFootprintArea</th>\n",
       "      <td>0.027718</td>\n",
       "      <td>0.006706</td>\n",
       "      <td>0.006058</td>\n",
       "      <td>0.013494</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>AvBlockFootprintArea</th>\n",
       "      <td>0.010309</td>\n",
       "      <td>0.011092</td>\n",
       "      <td>0.014500</td>\n",
       "      <td>0.011967</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total</th>\n",
       "      <td>0.401976</td>\n",
       "      <td>0.507942</td>\n",
       "      <td>0.550267</td>\n",
       "      <td>0.486728</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Country                                  Netherlands    France     Spain  \\\n",
       "feature                                                                    \n",
       "distance_to_closest_street                  0.032212  0.078079  0.022058   \n",
       "lat                                         0.022805  0.043192  0.062895   \n",
       "FootprintArea                               0.011029  0.057100  0.050449   \n",
       "StdBlockFootprintArea                       0.026739  0.013229  0.077640   \n",
       "buildings_within_buffer_100                 0.025147  0.037694  0.042311   \n",
       "street_betweeness_global_closest_street     0.031033  0.050493  0.018063   \n",
       "SharedWallLength                            0.017817  0.011597  0.058996   \n",
       "total_ft_area_within_buffer_500             0.028431  0.016130  0.028131   \n",
       "lon                                         0.010297  0.017277  0.036905   \n",
       "std_elongation_within_buffer_100            0.041104  0.004552  0.014099   \n",
       "street_width_av_closest_street              0.007405  0.022217  0.027337   \n",
       "av_footprint_area_within_buffer_100         0.015932  0.031816  0.007144   \n",
       "Phi                                         0.017194  0.032414  0.005182   \n",
       "Convexity                                   0.020747  0.007709  0.022656   \n",
       "total_ft_area_within_buffer_100             0.017750  0.010914  0.019386   \n",
       "std_convexity_within_buffer_100             0.014546  0.013240  0.016621   \n",
       "av_convexity_within_buffer_100              0.011077  0.021118  0.011193   \n",
       "BlockConvexity                              0.012685  0.021373  0.008645   \n",
       "BlockTotalFootprintArea                     0.027718  0.006706  0.006058   \n",
       "AvBlockFootprintArea                        0.010309  0.011092  0.014500   \n",
       "Total                                       0.401976  0.507942  0.550267   \n",
       "\n",
       "Country                                       All  \n",
       "feature                                            \n",
       "distance_to_closest_street               0.044116  \n",
       "lat                                      0.042964  \n",
       "FootprintArea                            0.039526  \n",
       "StdBlockFootprintArea                    0.039203  \n",
       "buildings_within_buffer_100              0.035051  \n",
       "street_betweeness_global_closest_street  0.033196  \n",
       "SharedWallLength                         0.029470  \n",
       "total_ft_area_within_buffer_500          0.024231  \n",
       "lon                                      0.021493  \n",
       "std_elongation_within_buffer_100         0.019918  \n",
       "street_width_av_closest_street           0.018986  \n",
       "av_footprint_area_within_buffer_100      0.018297  \n",
       "Phi                                      0.018263  \n",
       "Convexity                                0.017037  \n",
       "total_ft_area_within_buffer_100          0.016017  \n",
       "std_convexity_within_buffer_100          0.014802  \n",
       "av_convexity_within_buffer_100           0.014463  \n",
       "BlockConvexity                           0.014234  \n",
       "BlockTotalFootprintArea                  0.013494  \n",
       "AvBlockFootprintArea                     0.011967  \n",
       "Total                                    0.486728  "
      ]
     },
     "execution_count": 146,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_imp = df.rename(columns={'level_0': 'Country'})\n",
    "pivot_table = df_imp.pivot_table(values=['normalized_importance'], index=['feature'], columns=['Country'], margins=True)\n",
    "pivot_table.columns = pivot_table.columns.droplevel(0) # drop \"normalized_importance\" multi-index column\n",
    "\n",
    "new_cols = pivot_table.columns.reindex(['Netherlands', 'France', 'Spain', 'All'])\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "pivot_table = pivot_table.sort_values('All', ascending=False).head(20)\n",
    "\n",
    "pivot_table.loc[\"Total\"] = pivot_table.sum()\n",
    "pivot_table.to_latex('../figures/feature-importance.tex', index=True, float_format='%.4f', caption='Feature importance based on SHAP-values', label='table:fts-importance')\n",
    "pivot_table"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Figure: Feature values of time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {},
   "outputs": [],
   "source": [
    "# top 9 features\n",
    "fts = [\n",
    "    'distance_to_closest_street',\n",
    "    'FootprintArea',\n",
    "    'StdBlockFootprintArea',\n",
    "    'buildings_within_buffer_100',\n",
    "    'street_betweeness_global_closest_street',\n",
    "    'SharedWallLength',\n",
    "    'total_ft_area_within_buffer_500',\n",
    "    'std_elongation_within_buffer_100',\n",
    "    'street_width_av_closest_street',\n",
    "]\n",
    "# fts = dataset.FEATURES\n",
    "predictors = {'France': regressor_fts_importance_fr, 'Spain': regressor_fts_importance_es, 'Netherlands': regressor_fts_importance_nl}\n",
    "\n",
    "dfs = []\n",
    "for c, p in predictors.items():\n",
    "    df_country_fts = p.X_test[fts]\n",
    "    df_country_fts['country'] = c\n",
    "    df_country_fts = pd.concat([df_country_fts, p.y_test], axis=1, join=\"inner\")\n",
    "    dfs.append(df_country_fts)\n",
    "\n",
    "df_fts = pd.concat(dfs, axis=0)\n",
    "\n",
    "%matplotlib qt\n",
    "visualizations.plot_feature_over_time(df_fts, overlap_coef=overlap_coef_df, feature_selection=fts, round=3)\n",
    "plt.savefig('../figures/features-over-time-std.svg')\n",
    "\n",
    "visualizations.plot_feature_over_time(df_fts, feature_selection=fts, round=3)\n",
    "plt.savefig('../figures/features-over-time.svg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "87"
      ]
     },
     "execution_count": 135,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_fts_std = df_fts.groupby(['country'])[fts].agg(['std']).transpose()\n",
    "len(df_fts_std[(df_fts_std['Spain'] > df_fts_std['France']) & (df_fts_std['Spain'] > df_fts_std['Netherlands'])])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7143559605718741\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>FootprintArea</th>\n",
       "      <th>SharedWallLength</th>\n",
       "      <th>std_elongation_within_buffer_100</th>\n",
       "      <th>street_betweeness_global_closest_street</th>\n",
       "      <th>total_ft_area_within_buffer_500</th>\n",
       "      <th>distance_to_closest_street</th>\n",
       "      <th>buildings_within_buffer_100</th>\n",
       "      <th>street_width_av_closest_street</th>\n",
       "      <th>StdBlockFootprintArea</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1899.0</td>\n",
       "      <td>0.784955</td>\n",
       "      <td>0.849275</td>\n",
       "      <td>0.816711</td>\n",
       "      <td>0.865503</td>\n",
       "      <td>0.865064</td>\n",
       "      <td>0.929909</td>\n",
       "      <td>0.849391</td>\n",
       "      <td>0.769856</td>\n",
       "      <td>0.629146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1902.0</td>\n",
       "      <td>0.557458</td>\n",
       "      <td>0.825613</td>\n",
       "      <td>0.836703</td>\n",
       "      <td>0.788703</td>\n",
       "      <td>0.740010</td>\n",
       "      <td>0.934509</td>\n",
       "      <td>0.801855</td>\n",
       "      <td>0.823566</td>\n",
       "      <td>0.876653</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1905.0</td>\n",
       "      <td>0.581746</td>\n",
       "      <td>0.797945</td>\n",
       "      <td>0.842918</td>\n",
       "      <td>0.815813</td>\n",
       "      <td>0.656676</td>\n",
       "      <td>0.784858</td>\n",
       "      <td>0.731498</td>\n",
       "      <td>0.805743</td>\n",
       "      <td>0.589604</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1908.0</td>\n",
       "      <td>0.724559</td>\n",
       "      <td>0.829633</td>\n",
       "      <td>0.840468</td>\n",
       "      <td>0.826677</td>\n",
       "      <td>0.807269</td>\n",
       "      <td>0.601802</td>\n",
       "      <td>0.866290</td>\n",
       "      <td>0.765878</td>\n",
       "      <td>0.808649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1911.0</td>\n",
       "      <td>0.572721</td>\n",
       "      <td>0.798108</td>\n",
       "      <td>0.805014</td>\n",
       "      <td>0.883568</td>\n",
       "      <td>0.813586</td>\n",
       "      <td>0.948428</td>\n",
       "      <td>0.836194</td>\n",
       "      <td>0.738001</td>\n",
       "      <td>0.790463</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1914.0</td>\n",
       "      <td>0.731611</td>\n",
       "      <td>0.821895</td>\n",
       "      <td>0.831207</td>\n",
       "      <td>0.879028</td>\n",
       "      <td>0.722599</td>\n",
       "      <td>0.914385</td>\n",
       "      <td>0.827756</td>\n",
       "      <td>0.755318</td>\n",
       "      <td>0.889048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>1917.0</td>\n",
       "      <td>0.612407</td>\n",
       "      <td>0.791690</td>\n",
       "      <td>0.694483</td>\n",
       "      <td>0.810520</td>\n",
       "      <td>0.708999</td>\n",
       "      <td>0.948675</td>\n",
       "      <td>0.705046</td>\n",
       "      <td>0.770619</td>\n",
       "      <td>0.596889</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>1920.0</td>\n",
       "      <td>0.685530</td>\n",
       "      <td>0.813832</td>\n",
       "      <td>0.760950</td>\n",
       "      <td>0.832897</td>\n",
       "      <td>0.781262</td>\n",
       "      <td>0.769293</td>\n",
       "      <td>0.745090</td>\n",
       "      <td>0.759993</td>\n",
       "      <td>0.436327</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1923.0</td>\n",
       "      <td>0.731716</td>\n",
       "      <td>0.756255</td>\n",
       "      <td>0.744713</td>\n",
       "      <td>0.865325</td>\n",
       "      <td>0.614137</td>\n",
       "      <td>0.927059</td>\n",
       "      <td>0.678219</td>\n",
       "      <td>0.759007</td>\n",
       "      <td>0.354402</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1926.0</td>\n",
       "      <td>0.772947</td>\n",
       "      <td>0.748174</td>\n",
       "      <td>0.742286</td>\n",
       "      <td>0.853753</td>\n",
       "      <td>0.702268</td>\n",
       "      <td>0.769682</td>\n",
       "      <td>0.764737</td>\n",
       "      <td>0.745091</td>\n",
       "      <td>0.331789</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>1929.0</td>\n",
       "      <td>0.639662</td>\n",
       "      <td>0.776147</td>\n",
       "      <td>0.772467</td>\n",
       "      <td>0.895073</td>\n",
       "      <td>0.782093</td>\n",
       "      <td>0.873453</td>\n",
       "      <td>0.766855</td>\n",
       "      <td>0.731723</td>\n",
       "      <td>0.491797</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>1932.0</td>\n",
       "      <td>0.846622</td>\n",
       "      <td>0.726750</td>\n",
       "      <td>0.704446</td>\n",
       "      <td>0.833444</td>\n",
       "      <td>0.687967</td>\n",
       "      <td>0.658575</td>\n",
       "      <td>0.702003</td>\n",
       "      <td>0.743597</td>\n",
       "      <td>0.663137</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>1935.0</td>\n",
       "      <td>0.720769</td>\n",
       "      <td>0.735279</td>\n",
       "      <td>0.758798</td>\n",
       "      <td>0.909257</td>\n",
       "      <td>0.728608</td>\n",
       "      <td>0.744381</td>\n",
       "      <td>0.748981</td>\n",
       "      <td>0.701734</td>\n",
       "      <td>0.741002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>1938.0</td>\n",
       "      <td>0.738095</td>\n",
       "      <td>0.743574</td>\n",
       "      <td>0.745748</td>\n",
       "      <td>0.820400</td>\n",
       "      <td>0.744057</td>\n",
       "      <td>0.677359</td>\n",
       "      <td>0.725422</td>\n",
       "      <td>0.709708</td>\n",
       "      <td>0.631773</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>1941.0</td>\n",
       "      <td>0.953449</td>\n",
       "      <td>0.776353</td>\n",
       "      <td>0.636342</td>\n",
       "      <td>0.793559</td>\n",
       "      <td>0.724588</td>\n",
       "      <td>0.722433</td>\n",
       "      <td>0.637035</td>\n",
       "      <td>0.770392</td>\n",
       "      <td>0.733249</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>1944.0</td>\n",
       "      <td>0.512494</td>\n",
       "      <td>0.765054</td>\n",
       "      <td>0.799930</td>\n",
       "      <td>0.859153</td>\n",
       "      <td>0.729302</td>\n",
       "      <td>0.767354</td>\n",
       "      <td>0.736121</td>\n",
       "      <td>0.686617</td>\n",
       "      <td>0.328170</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>1947.0</td>\n",
       "      <td>0.756202</td>\n",
       "      <td>0.736237</td>\n",
       "      <td>0.786910</td>\n",
       "      <td>0.775478</td>\n",
       "      <td>0.698278</td>\n",
       "      <td>0.847311</td>\n",
       "      <td>0.786613</td>\n",
       "      <td>0.729960</td>\n",
       "      <td>0.770015</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>1950.0</td>\n",
       "      <td>0.809241</td>\n",
       "      <td>0.764656</td>\n",
       "      <td>0.752484</td>\n",
       "      <td>0.811137</td>\n",
       "      <td>0.727555</td>\n",
       "      <td>0.793389</td>\n",
       "      <td>0.795365</td>\n",
       "      <td>0.738785</td>\n",
       "      <td>0.719448</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>1953.0</td>\n",
       "      <td>0.577072</td>\n",
       "      <td>0.717035</td>\n",
       "      <td>0.716812</td>\n",
       "      <td>0.809539</td>\n",
       "      <td>0.662900</td>\n",
       "      <td>0.942719</td>\n",
       "      <td>0.719291</td>\n",
       "      <td>0.726421</td>\n",
       "      <td>0.715406</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>1956.0</td>\n",
       "      <td>0.508206</td>\n",
       "      <td>0.697399</td>\n",
       "      <td>0.699648</td>\n",
       "      <td>0.817074</td>\n",
       "      <td>0.659954</td>\n",
       "      <td>0.735638</td>\n",
       "      <td>0.713687</td>\n",
       "      <td>0.714098</td>\n",
       "      <td>0.478593</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>1959.0</td>\n",
       "      <td>0.552024</td>\n",
       "      <td>0.742202</td>\n",
       "      <td>0.726926</td>\n",
       "      <td>0.813211</td>\n",
       "      <td>0.675694</td>\n",
       "      <td>0.739497</td>\n",
       "      <td>0.754450</td>\n",
       "      <td>0.724928</td>\n",
       "      <td>0.392785</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>1962.0</td>\n",
       "      <td>0.703352</td>\n",
       "      <td>0.666854</td>\n",
       "      <td>0.693206</td>\n",
       "      <td>0.822172</td>\n",
       "      <td>0.571279</td>\n",
       "      <td>0.872405</td>\n",
       "      <td>0.688941</td>\n",
       "      <td>0.722862</td>\n",
       "      <td>0.346959</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>1965.0</td>\n",
       "      <td>0.634728</td>\n",
       "      <td>0.658433</td>\n",
       "      <td>0.706705</td>\n",
       "      <td>0.794805</td>\n",
       "      <td>0.578112</td>\n",
       "      <td>0.795751</td>\n",
       "      <td>0.640552</td>\n",
       "      <td>0.739236</td>\n",
       "      <td>0.260306</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>1968.0</td>\n",
       "      <td>0.636359</td>\n",
       "      <td>0.622721</td>\n",
       "      <td>0.701652</td>\n",
       "      <td>0.763324</td>\n",
       "      <td>0.549580</td>\n",
       "      <td>0.743385</td>\n",
       "      <td>0.591223</td>\n",
       "      <td>0.741003</td>\n",
       "      <td>0.389810</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>1971.0</td>\n",
       "      <td>0.444031</td>\n",
       "      <td>0.702020</td>\n",
       "      <td>0.723330</td>\n",
       "      <td>0.705149</td>\n",
       "      <td>0.588158</td>\n",
       "      <td>0.709711</td>\n",
       "      <td>0.605603</td>\n",
       "      <td>0.767302</td>\n",
       "      <td>0.359843</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>1974.0</td>\n",
       "      <td>0.671119</td>\n",
       "      <td>0.715740</td>\n",
       "      <td>0.737447</td>\n",
       "      <td>0.719474</td>\n",
       "      <td>0.594601</td>\n",
       "      <td>0.608361</td>\n",
       "      <td>0.611859</td>\n",
       "      <td>0.799909</td>\n",
       "      <td>0.351088</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>1977.0</td>\n",
       "      <td>0.411037</td>\n",
       "      <td>0.685622</td>\n",
       "      <td>0.798446</td>\n",
       "      <td>0.749075</td>\n",
       "      <td>0.566627</td>\n",
       "      <td>0.762330</td>\n",
       "      <td>0.615904</td>\n",
       "      <td>0.767704</td>\n",
       "      <td>0.263497</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>1980.0</td>\n",
       "      <td>0.427836</td>\n",
       "      <td>0.743931</td>\n",
       "      <td>0.785106</td>\n",
       "      <td>0.733297</td>\n",
       "      <td>0.619478</td>\n",
       "      <td>0.644178</td>\n",
       "      <td>0.626088</td>\n",
       "      <td>0.797203</td>\n",
       "      <td>0.380774</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>1983.0</td>\n",
       "      <td>0.621507</td>\n",
       "      <td>0.753364</td>\n",
       "      <td>0.716460</td>\n",
       "      <td>0.723753</td>\n",
       "      <td>0.614806</td>\n",
       "      <td>0.667290</td>\n",
       "      <td>0.562923</td>\n",
       "      <td>0.804350</td>\n",
       "      <td>0.379177</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>1986.0</td>\n",
       "      <td>0.644856</td>\n",
       "      <td>0.794001</td>\n",
       "      <td>0.683459</td>\n",
       "      <td>0.723931</td>\n",
       "      <td>0.655711</td>\n",
       "      <td>0.691675</td>\n",
       "      <td>0.629410</td>\n",
       "      <td>0.827387</td>\n",
       "      <td>0.312250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>1989.0</td>\n",
       "      <td>0.794177</td>\n",
       "      <td>0.842535</td>\n",
       "      <td>0.657397</td>\n",
       "      <td>0.738467</td>\n",
       "      <td>0.692786</td>\n",
       "      <td>0.662325</td>\n",
       "      <td>0.686370</td>\n",
       "      <td>0.855342</td>\n",
       "      <td>0.556880</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>1992.0</td>\n",
       "      <td>0.764548</td>\n",
       "      <td>0.794400</td>\n",
       "      <td>0.661756</td>\n",
       "      <td>0.763918</td>\n",
       "      <td>0.647930</td>\n",
       "      <td>0.840922</td>\n",
       "      <td>0.717498</td>\n",
       "      <td>0.823865</td>\n",
       "      <td>0.530551</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>1995.0</td>\n",
       "      <td>0.568859</td>\n",
       "      <td>0.849201</td>\n",
       "      <td>0.678564</td>\n",
       "      <td>0.741854</td>\n",
       "      <td>0.686856</td>\n",
       "      <td>0.686937</td>\n",
       "      <td>0.699436</td>\n",
       "      <td>0.837012</td>\n",
       "      <td>0.525707</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>1998.0</td>\n",
       "      <td>0.516745</td>\n",
       "      <td>0.811137</td>\n",
       "      <td>0.705217</td>\n",
       "      <td>0.761434</td>\n",
       "      <td>0.675388</td>\n",
       "      <td>0.669560</td>\n",
       "      <td>0.669453</td>\n",
       "      <td>0.829650</td>\n",
       "      <td>0.381751</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>2001.0</td>\n",
       "      <td>0.717251</td>\n",
       "      <td>0.858319</td>\n",
       "      <td>0.710814</td>\n",
       "      <td>0.752540</td>\n",
       "      <td>0.680340</td>\n",
       "      <td>0.651962</td>\n",
       "      <td>0.639358</td>\n",
       "      <td>0.842852</td>\n",
       "      <td>0.536923</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>2004.0</td>\n",
       "      <td>0.502161</td>\n",
       "      <td>0.888112</td>\n",
       "      <td>0.736881</td>\n",
       "      <td>0.772565</td>\n",
       "      <td>0.677817</td>\n",
       "      <td>0.710728</td>\n",
       "      <td>0.658398</td>\n",
       "      <td>0.853770</td>\n",
       "      <td>0.538474</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>2007.0</td>\n",
       "      <td>0.469838</td>\n",
       "      <td>0.905035</td>\n",
       "      <td>0.744388</td>\n",
       "      <td>0.790669</td>\n",
       "      <td>0.690907</td>\n",
       "      <td>0.657631</td>\n",
       "      <td>0.675909</td>\n",
       "      <td>0.865612</td>\n",
       "      <td>0.430930</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>2010.0</td>\n",
       "      <td>0.575308</td>\n",
       "      <td>0.925259</td>\n",
       "      <td>0.768863</td>\n",
       "      <td>0.819859</td>\n",
       "      <td>0.708589</td>\n",
       "      <td>0.693812</td>\n",
       "      <td>0.691186</td>\n",
       "      <td>0.861732</td>\n",
       "      <td>0.551564</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>2013.0</td>\n",
       "      <td>0.741024</td>\n",
       "      <td>0.909849</td>\n",
       "      <td>0.759919</td>\n",
       "      <td>0.800826</td>\n",
       "      <td>0.734714</td>\n",
       "      <td>0.686664</td>\n",
       "      <td>0.694282</td>\n",
       "      <td>0.861273</td>\n",
       "      <td>0.509539</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>2016.0</td>\n",
       "      <td>0.625814</td>\n",
       "      <td>0.931051</td>\n",
       "      <td>0.737213</td>\n",
       "      <td>0.762924</td>\n",
       "      <td>0.723832</td>\n",
       "      <td>0.593120</td>\n",
       "      <td>0.720130</td>\n",
       "      <td>0.877310</td>\n",
       "      <td>0.428446</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total</th>\n",
       "      <td>1957.5</td>\n",
       "      <td>0.646001</td>\n",
       "      <td>0.781767</td>\n",
       "      <td>0.743070</td>\n",
       "      <td>0.799979</td>\n",
       "      <td>0.687260</td>\n",
       "      <td>0.759486</td>\n",
       "      <td>0.707911</td>\n",
       "      <td>0.778660</td>\n",
       "      <td>0.525070</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          age  FootprintArea  SharedWallLength  \\\n",
       "0      1899.0       0.784955          0.849275   \n",
       "1      1902.0       0.557458          0.825613   \n",
       "2      1905.0       0.581746          0.797945   \n",
       "3      1908.0       0.724559          0.829633   \n",
       "4      1911.0       0.572721          0.798108   \n",
       "5      1914.0       0.731611          0.821895   \n",
       "6      1917.0       0.612407          0.791690   \n",
       "7      1920.0       0.685530          0.813832   \n",
       "8      1923.0       0.731716          0.756255   \n",
       "9      1926.0       0.772947          0.748174   \n",
       "10     1929.0       0.639662          0.776147   \n",
       "11     1932.0       0.846622          0.726750   \n",
       "12     1935.0       0.720769          0.735279   \n",
       "13     1938.0       0.738095          0.743574   \n",
       "14     1941.0       0.953449          0.776353   \n",
       "15     1944.0       0.512494          0.765054   \n",
       "16     1947.0       0.756202          0.736237   \n",
       "17     1950.0       0.809241          0.764656   \n",
       "18     1953.0       0.577072          0.717035   \n",
       "19     1956.0       0.508206          0.697399   \n",
       "20     1959.0       0.552024          0.742202   \n",
       "21     1962.0       0.703352          0.666854   \n",
       "22     1965.0       0.634728          0.658433   \n",
       "23     1968.0       0.636359          0.622721   \n",
       "24     1971.0       0.444031          0.702020   \n",
       "25     1974.0       0.671119          0.715740   \n",
       "26     1977.0       0.411037          0.685622   \n",
       "27     1980.0       0.427836          0.743931   \n",
       "28     1983.0       0.621507          0.753364   \n",
       "29     1986.0       0.644856          0.794001   \n",
       "30     1989.0       0.794177          0.842535   \n",
       "31     1992.0       0.764548          0.794400   \n",
       "32     1995.0       0.568859          0.849201   \n",
       "33     1998.0       0.516745          0.811137   \n",
       "34     2001.0       0.717251          0.858319   \n",
       "35     2004.0       0.502161          0.888112   \n",
       "36     2007.0       0.469838          0.905035   \n",
       "37     2010.0       0.575308          0.925259   \n",
       "38     2013.0       0.741024          0.909849   \n",
       "39     2016.0       0.625814          0.931051   \n",
       "Total  1957.5       0.646001          0.781767   \n",
       "\n",
       "       std_elongation_within_buffer_100  \\\n",
       "0                              0.816711   \n",
       "1                              0.836703   \n",
       "2                              0.842918   \n",
       "3                              0.840468   \n",
       "4                              0.805014   \n",
       "5                              0.831207   \n",
       "6                              0.694483   \n",
       "7                              0.760950   \n",
       "8                              0.744713   \n",
       "9                              0.742286   \n",
       "10                             0.772467   \n",
       "11                             0.704446   \n",
       "12                             0.758798   \n",
       "13                             0.745748   \n",
       "14                             0.636342   \n",
       "15                             0.799930   \n",
       "16                             0.786910   \n",
       "17                             0.752484   \n",
       "18                             0.716812   \n",
       "19                             0.699648   \n",
       "20                             0.726926   \n",
       "21                             0.693206   \n",
       "22                             0.706705   \n",
       "23                             0.701652   \n",
       "24                             0.723330   \n",
       "25                             0.737447   \n",
       "26                             0.798446   \n",
       "27                             0.785106   \n",
       "28                             0.716460   \n",
       "29                             0.683459   \n",
       "30                             0.657397   \n",
       "31                             0.661756   \n",
       "32                             0.678564   \n",
       "33                             0.705217   \n",
       "34                             0.710814   \n",
       "35                             0.736881   \n",
       "36                             0.744388   \n",
       "37                             0.768863   \n",
       "38                             0.759919   \n",
       "39                             0.737213   \n",
       "Total                          0.743070   \n",
       "\n",
       "       street_betweeness_global_closest_street  \\\n",
       "0                                     0.865503   \n",
       "1                                     0.788703   \n",
       "2                                     0.815813   \n",
       "3                                     0.826677   \n",
       "4                                     0.883568   \n",
       "5                                     0.879028   \n",
       "6                                     0.810520   \n",
       "7                                     0.832897   \n",
       "8                                     0.865325   \n",
       "9                                     0.853753   \n",
       "10                                    0.895073   \n",
       "11                                    0.833444   \n",
       "12                                    0.909257   \n",
       "13                                    0.820400   \n",
       "14                                    0.793559   \n",
       "15                                    0.859153   \n",
       "16                                    0.775478   \n",
       "17                                    0.811137   \n",
       "18                                    0.809539   \n",
       "19                                    0.817074   \n",
       "20                                    0.813211   \n",
       "21                                    0.822172   \n",
       "22                                    0.794805   \n",
       "23                                    0.763324   \n",
       "24                                    0.705149   \n",
       "25                                    0.719474   \n",
       "26                                    0.749075   \n",
       "27                                    0.733297   \n",
       "28                                    0.723753   \n",
       "29                                    0.723931   \n",
       "30                                    0.738467   \n",
       "31                                    0.763918   \n",
       "32                                    0.741854   \n",
       "33                                    0.761434   \n",
       "34                                    0.752540   \n",
       "35                                    0.772565   \n",
       "36                                    0.790669   \n",
       "37                                    0.819859   \n",
       "38                                    0.800826   \n",
       "39                                    0.762924   \n",
       "Total                                 0.799979   \n",
       "\n",
       "       total_ft_area_within_buffer_500  distance_to_closest_street  \\\n",
       "0                             0.865064                    0.929909   \n",
       "1                             0.740010                    0.934509   \n",
       "2                             0.656676                    0.784858   \n",
       "3                             0.807269                    0.601802   \n",
       "4                             0.813586                    0.948428   \n",
       "5                             0.722599                    0.914385   \n",
       "6                             0.708999                    0.948675   \n",
       "7                             0.781262                    0.769293   \n",
       "8                             0.614137                    0.927059   \n",
       "9                             0.702268                    0.769682   \n",
       "10                            0.782093                    0.873453   \n",
       "11                            0.687967                    0.658575   \n",
       "12                            0.728608                    0.744381   \n",
       "13                            0.744057                    0.677359   \n",
       "14                            0.724588                    0.722433   \n",
       "15                            0.729302                    0.767354   \n",
       "16                            0.698278                    0.847311   \n",
       "17                            0.727555                    0.793389   \n",
       "18                            0.662900                    0.942719   \n",
       "19                            0.659954                    0.735638   \n",
       "20                            0.675694                    0.739497   \n",
       "21                            0.571279                    0.872405   \n",
       "22                            0.578112                    0.795751   \n",
       "23                            0.549580                    0.743385   \n",
       "24                            0.588158                    0.709711   \n",
       "25                            0.594601                    0.608361   \n",
       "26                            0.566627                    0.762330   \n",
       "27                            0.619478                    0.644178   \n",
       "28                            0.614806                    0.667290   \n",
       "29                            0.655711                    0.691675   \n",
       "30                            0.692786                    0.662325   \n",
       "31                            0.647930                    0.840922   \n",
       "32                            0.686856                    0.686937   \n",
       "33                            0.675388                    0.669560   \n",
       "34                            0.680340                    0.651962   \n",
       "35                            0.677817                    0.710728   \n",
       "36                            0.690907                    0.657631   \n",
       "37                            0.708589                    0.693812   \n",
       "38                            0.734714                    0.686664   \n",
       "39                            0.723832                    0.593120   \n",
       "Total                         0.687260                    0.759486   \n",
       "\n",
       "       buildings_within_buffer_100  street_width_av_closest_street  \\\n",
       "0                         0.849391                        0.769856   \n",
       "1                         0.801855                        0.823566   \n",
       "2                         0.731498                        0.805743   \n",
       "3                         0.866290                        0.765878   \n",
       "4                         0.836194                        0.738001   \n",
       "5                         0.827756                        0.755318   \n",
       "6                         0.705046                        0.770619   \n",
       "7                         0.745090                        0.759993   \n",
       "8                         0.678219                        0.759007   \n",
       "9                         0.764737                        0.745091   \n",
       "10                        0.766855                        0.731723   \n",
       "11                        0.702003                        0.743597   \n",
       "12                        0.748981                        0.701734   \n",
       "13                        0.725422                        0.709708   \n",
       "14                        0.637035                        0.770392   \n",
       "15                        0.736121                        0.686617   \n",
       "16                        0.786613                        0.729960   \n",
       "17                        0.795365                        0.738785   \n",
       "18                        0.719291                        0.726421   \n",
       "19                        0.713687                        0.714098   \n",
       "20                        0.754450                        0.724928   \n",
       "21                        0.688941                        0.722862   \n",
       "22                        0.640552                        0.739236   \n",
       "23                        0.591223                        0.741003   \n",
       "24                        0.605603                        0.767302   \n",
       "25                        0.611859                        0.799909   \n",
       "26                        0.615904                        0.767704   \n",
       "27                        0.626088                        0.797203   \n",
       "28                        0.562923                        0.804350   \n",
       "29                        0.629410                        0.827387   \n",
       "30                        0.686370                        0.855342   \n",
       "31                        0.717498                        0.823865   \n",
       "32                        0.699436                        0.837012   \n",
       "33                        0.669453                        0.829650   \n",
       "34                        0.639358                        0.842852   \n",
       "35                        0.658398                        0.853770   \n",
       "36                        0.675909                        0.865612   \n",
       "37                        0.691186                        0.861732   \n",
       "38                        0.694282                        0.861273   \n",
       "39                        0.720130                        0.877310   \n",
       "Total                     0.707911                        0.778660   \n",
       "\n",
       "       StdBlockFootprintArea  \n",
       "0                   0.629146  \n",
       "1                   0.876653  \n",
       "2                   0.589604  \n",
       "3                   0.808649  \n",
       "4                   0.790463  \n",
       "5                   0.889048  \n",
       "6                   0.596889  \n",
       "7                   0.436327  \n",
       "8                   0.354402  \n",
       "9                   0.331789  \n",
       "10                  0.491797  \n",
       "11                  0.663137  \n",
       "12                  0.741002  \n",
       "13                  0.631773  \n",
       "14                  0.733249  \n",
       "15                  0.328170  \n",
       "16                  0.770015  \n",
       "17                  0.719448  \n",
       "18                  0.715406  \n",
       "19                  0.478593  \n",
       "20                  0.392785  \n",
       "21                  0.346959  \n",
       "22                  0.260306  \n",
       "23                  0.389810  \n",
       "24                  0.359843  \n",
       "25                  0.351088  \n",
       "26                  0.263497  \n",
       "27                  0.380774  \n",
       "28                  0.379177  \n",
       "29                  0.312250  \n",
       "30                  0.556880  \n",
       "31                  0.530551  \n",
       "32                  0.525707  \n",
       "33                  0.381751  \n",
       "34                  0.536923  \n",
       "35                  0.538474  \n",
       "36                  0.430930  \n",
       "37                  0.551564  \n",
       "38                  0.509539  \n",
       "39                  0.428446  \n",
       "Total               0.525070  "
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from statistics import NormalDist\n",
    "\n",
    "def calc_dis_overlap_coef(df):\n",
    "    coef = []\n",
    "    fts = set(df.columns.get_level_values(0).drop(['country', 'rounded_age']))\n",
    "    for ft in fts:\n",
    "        try:\n",
    "            mean_fr, std_fr = df[df['country'] == 'France'][ft].values[0]\n",
    "            mean_es, std_es = df[df['country'] == 'Spain'][ft].values[0]\n",
    "            mean_nl, std_nl = df[df['country'] == 'Netherlands'][ft].values[0]\n",
    "            coef_fr_es = NormalDist(mu=mean_fr, sigma=std_fr).overlap(NormalDist(mu=mean_es, sigma=std_es))\n",
    "            coef_fr_nl = NormalDist(mu=mean_fr, sigma=std_fr).overlap(NormalDist(mu=mean_nl, sigma=std_nl))\n",
    "            coef_es_nl = NormalDist(mu=mean_es, sigma=std_es).overlap(NormalDist(mu=mean_nl, sigma=std_nl))\n",
    "            coef.append(np.mean([coef_fr_es, coef_fr_nl, coef_es_nl]))\n",
    "        except IndexError:\n",
    "            coef.append(np.nan)\n",
    "\n",
    "    return pd.Series(coef, index=fts)\n",
    "\n",
    "\n",
    "df_fts['rounded_age'] = utils.custom_round(df_fts[dataset.AGE_ATTRIBUTE], base=3)\n",
    "df = df_fts.groupby(['country', 'rounded_age'])[fts].agg(['mean', 'std']).reset_index()\n",
    "overlap_coef_df = df.groupby('rounded_age').apply(calc_dis_overlap_coef).reset_index().rename(columns={'rounded_age': 'age'}).dropna()\n",
    "overlap_coef_df.loc['Total']= overlap_coef_df.mean()\n",
    "print(overlap_coef_df.loc['Total'][fts].mean())\n",
    "overlap_coef_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Figure: Feature distribution differences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 439,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    }
   ],
   "source": [
    "# fts = dataset.SELECTED_FEATURES[:4]\n",
    "# fts = ['BlockElongation', 'buildings_within_buffer_100', 'Corners', 'street_width_av_closest_street', 'FootprintArea', 'total_ft_area_within_buffer_100']\n",
    "# fts = top_10_features\n",
    "fts = [\n",
    "    'distance_to_closest_street',\n",
    "    'FootprintArea',\n",
    "    'StdBlockFootprintArea',\n",
    "    'buildings_within_buffer_100',\n",
    "    'street_betweeness_global_closest_street',\n",
    "    'SharedWallLength',\n",
    "    'total_ft_area_within_buffer_500',\n",
    "    'std_elongation_within_buffer_100',\n",
    "]\n",
    "predictors = {'France': regressor_fts_importance_fr, 'Spain': regressor_fts_importance_es, 'Netherlands': regressor_fts_importance_nl}\n",
    "dfs = []\n",
    "\n",
    "for c, p in predictors.items():\n",
    "    df_country_fts = p.X_test[fts]\n",
    "    df_country_fts['Country'] = c\n",
    "    dfs.append(df_country_fts)\n",
    "\n",
    "df_fts = pd.concat(dfs, axis=0)\n",
    "\n",
    "# fts.remove('lat')\n",
    "# fts.remove('lon')\n",
    "\n",
    "# _, axis = plt.subplots(nrows=len(fts), ncols=1, figsize=(30, 20), constrained_layout=True)\n",
    "axis = utils.grid_subplot(len(fts), n_cols=2)\n",
    "for idx, ft in enumerate(fts):\n",
    "    # min = df_fts[ft].quantile(0.05)\n",
    "    # max = df_fts[ft].quantile(0.95)\n",
    "    iqr = df_fts[ft].quantile(0.75) - df_fts[ft].quantile(0.25)\n",
    "    min = df_fts[ft].mean() - iqr\n",
    "    max = df_fts[ft].mean() + iqr\n",
    "\n",
    "    colors = ['lightseagreen', 'plum', 'darksalmon']\n",
    "    sns.kdeplot(x=ft, data=df_fts, hue='Country', shade=True, clip=(min, max), palette=colors, common_norm=False, ax=axis[idx], legend=False)\n",
    "    axis[idx].set_axis_off()\n",
    "    axis[idx].set_title(ft, y=0, pad=-15, verticalalignment='top')\n",
    "\n",
    "axis[0].legend(loc='lower center', bbox_to_anchor=(0.5, 1.1),\n",
    "          fancybox=True, ncol=3, labels=['Netherlands', 'Spain', 'France']).set_draggable(state=True)\n",
    "\n",
    "plt.show()\n",
    "plt.savefig('../figures/feature-distributions-2.svg')\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Other building attributes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 465,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Netherlands</th>\n",
       "      <th colspan=\"3\" halign=\"left\">France</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Spain</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>residential</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>16.956241</td>\n",
       "      <td>23.261603</td>\n",
       "      <td>0.436896</td>\n",
       "      <td>19.389091</td>\n",
       "      <td>26.287121</td>\n",
       "      <td>0.340885</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>all</th>\n",
       "      <td>13.895980</td>\n",
       "      <td>19.422970</td>\n",
       "      <td>0.464333</td>\n",
       "      <td>17.585913</td>\n",
       "      <td>23.980237</td>\n",
       "      <td>0.428436</td>\n",
       "      <td>18.669917</td>\n",
       "      <td>25.663298</td>\n",
       "      <td>0.374679</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>height</th>\n",
       "      <td>13.903681</td>\n",
       "      <td>19.426731</td>\n",
       "      <td>0.464126</td>\n",
       "      <td>17.747595</td>\n",
       "      <td>24.187547</td>\n",
       "      <td>0.418511</td>\n",
       "      <td>18.838526</td>\n",
       "      <td>25.854203</td>\n",
       "      <td>0.365341</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>type</th>\n",
       "      <td>14.303375</td>\n",
       "      <td>19.862269</td>\n",
       "      <td>0.439828</td>\n",
       "      <td>17.821466</td>\n",
       "      <td>24.218596</td>\n",
       "      <td>0.417017</td>\n",
       "      <td>19.088726</td>\n",
       "      <td>26.061900</td>\n",
       "      <td>0.355103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>baseline</th>\n",
       "      <td>14.310712</td>\n",
       "      <td>19.865808</td>\n",
       "      <td>0.439629</td>\n",
       "      <td>17.999451</td>\n",
       "      <td>24.436980</td>\n",
       "      <td>0.406456</td>\n",
       "      <td>19.294297</td>\n",
       "      <td>26.288578</td>\n",
       "      <td>0.343836</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Country     Netherlands                          France                       \\\n",
       "                    MAE       RMSE        R2        MAE       RMSE        R2   \n",
       "Model                                                                          \n",
       "residential         NaN        NaN       NaN  16.956241  23.261603  0.436896   \n",
       "all           13.895980  19.422970  0.464333  17.585913  23.980237  0.428436   \n",
       "height        13.903681  19.426731  0.464126  17.747595  24.187547  0.418511   \n",
       "type          14.303375  19.862269  0.439828  17.821466  24.218596  0.417017   \n",
       "baseline      14.310712  19.865808  0.439629  17.999451  24.436980  0.406456   \n",
       "\n",
       "Country          Spain                       \n",
       "                   MAE       RMSE        R2  \n",
       "Model                                        \n",
       "residential  19.389091  26.287121  0.340885  \n",
       "all          18.669917  25.663298  0.374679  \n",
       "height       18.838526  25.854203  0.365341  \n",
       "type         19.088726  26.061900  0.355103  \n",
       "baseline     19.294297  26.288578  0.343836  "
      ]
     },
     "execution_count": 465,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "add_features_results = pd.read_csv('../results/comparison-results-additional-features.csv')\n",
    "add_features_results['Country'] = add_features_results['name'].str.split('_').str[0]\n",
    "add_features_results['Model'] = add_features_results['name'].str.split('_').str[1]\n",
    "add_features_results = add_features_results[~add_features_results['Model'].str.contains('features')]\n",
    "# add_features_results = add_features_results[~add_features_results['Model'].str.contains('residential')]\n",
    "add_features_results.sort_values(by=['Country', 'R2'])\n",
    "\n",
    "pivot_table = add_features_results.pivot_table(values=['MAE', 'RMSE', 'R2'], index=['Model'], columns=['Country'], margins=False)\n",
    "pivot_table = pivot_table.swaplevel(0, 1, axis=1)\n",
    "pivot_table = pivot_table.sort_index(axis=1, level=0)\n",
    "\n",
    "new_cols = pivot_table.columns.reindex(['Netherlands', 'France', 'Spain', 'All'], level=0)\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "new_cols = pivot_table.columns.reindex(['MAE', 'RMSE', 'R2'], level=1)\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "\n",
    "pivot_table = pivot_table.sort_values(by=('France', 'R2'), ascending=False)\n",
    "# pivot_table.to_latex('../figures/error-additional-attributes.tex', index=True, float_format='%.2f', caption='Prediction error for additional features', label='table:error-additional-attributes')\n",
    "pivot_table\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 302,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"3\" halign=\"left\">Netherlands</th>\n",
       "      <th colspan=\"3\" halign=\"left\">France</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Spain</th>\n",
       "      <th colspan=\"3\" halign=\"left\">All</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>R2</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Random cv</th>\n",
       "      <td>0.564887</td>\n",
       "      <td>1.096148</td>\n",
       "      <td>0.838229</td>\n",
       "      <td>1.168369</td>\n",
       "      <td>1.658524</td>\n",
       "      <td>0.560936</td>\n",
       "      <td>1.203284</td>\n",
       "      <td>2.023251</td>\n",
       "      <td>0.706021</td>\n",
       "      <td>1.084387</td>\n",
       "      <td>1.733827</td>\n",
       "      <td>0.670829</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block cv</th>\n",
       "      <td>0.663976</td>\n",
       "      <td>1.250968</td>\n",
       "      <td>0.789305</td>\n",
       "      <td>1.227446</td>\n",
       "      <td>1.749637</td>\n",
       "      <td>0.511370</td>\n",
       "      <td>1.284153</td>\n",
       "      <td>2.174640</td>\n",
       "      <td>0.660381</td>\n",
       "      <td>1.131874</td>\n",
       "      <td>1.824111</td>\n",
       "      <td>0.635655</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>0.706079</td>\n",
       "      <td>1.299744</td>\n",
       "      <td>0.772555</td>\n",
       "      <td>1.263358</td>\n",
       "      <td>1.800029</td>\n",
       "      <td>0.482819</td>\n",
       "      <td>1.362364</td>\n",
       "      <td>2.300560</td>\n",
       "      <td>0.619912</td>\n",
       "      <td>1.174134</td>\n",
       "      <td>1.895153</td>\n",
       "      <td>0.606723</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>0.734354</td>\n",
       "      <td>1.344791</td>\n",
       "      <td>0.756516</td>\n",
       "      <td>1.306926</td>\n",
       "      <td>1.856003</td>\n",
       "      <td>0.450153</td>\n",
       "      <td>1.432243</td>\n",
       "      <td>2.445343</td>\n",
       "      <td>0.570565</td>\n",
       "      <td>1.226016</td>\n",
       "      <td>1.992226</td>\n",
       "      <td>0.565403</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                Netherlands                        France                      \\\n",
       "                        MAE      RMSE        R2       MAE      RMSE        R2   \n",
       "Model                                                                           \n",
       "Random cv          0.564887  1.096148  0.838229  1.168369  1.658524  0.560936   \n",
       "Block cv           0.663976  1.250968  0.789305  1.227446  1.749637  0.511370   \n",
       "Neighborhood cv    0.706079  1.299744  0.772555  1.263358  1.800029  0.482819   \n",
       "City cv            0.734354  1.344791  0.756516  1.306926  1.856003  0.450153   \n",
       "\n",
       "                    Spain                           All                      \n",
       "                      MAE      RMSE        R2       MAE      RMSE        R2  \n",
       "Model                                                                        \n",
       "Random cv        1.203284  2.023251  0.706021  1.084387  1.733827  0.670829  \n",
       "Block cv         1.284153  2.174640  0.660381  1.131874  1.824111  0.635655  \n",
       "Neighborhood cv  1.362364  2.300560  0.619912  1.174134  1.895153  0.606723  \n",
       "City cv          1.432243  2.445343  0.570565  1.226016  1.992226  0.565403  "
      ]
     },
     "execution_count": 302,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "height_results = pd.read_csv('../results/results-height-country-comparison.csv')\n",
    "height_results['Country'] = height_results['name'].str.split('_').str[0]\n",
    "height_results['Model'] = height_results['name'].str.split('_').str[1]\n",
    "height_results['Model'] = height_results['Model'].str.replace('-', ' ').str.capitalize()\n",
    "\n",
    "pivot_table = height_results.pivot_table(values=['MAE', 'RMSE', 'R2'], index=['Model'], columns=['Country'], margins=False)\n",
    "pivot_table = pivot_table.swaplevel(0, 1, axis=1)\n",
    "pivot_table = pivot_table.sort_index(axis=1, level=0)\n",
    "\n",
    "pivot_table.columns = pivot_table.columns.rename(None, level=0) # drop Country cell\n",
    "\n",
    "new_cols = pivot_table.columns.reindex(['Netherlands', 'France', 'Spain', 'All'], level=0)\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "new_cols = pivot_table.columns.reindex(['MAE', 'RMSE', 'R2'], level=1)\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "\n",
    "pivot_table = pivot_table.sort_values(by=('Netherlands', 'R2'), ascending=False)\n",
    "pivot_table.to_latex('../figures/error-height-prediction.tex', index=True, float_format='%.2f', caption='Height prediction error', label='table:height-prediction-results')\n",
    "pivot_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"3\" halign=\"left\">France</th>\n",
       "      <th colspan=\"3\" halign=\"left\">Spain</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>MCC</th>\n",
       "      <th>F1</th>\n",
       "      <th>Recall residential</th>\n",
       "      <th>MCC</th>\n",
       "      <th>F1</th>\n",
       "      <th>Recall residential</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Random cv</th>\n",
       "      <td>0.605411</td>\n",
       "      <td>0.799586</td>\n",
       "      <td>0.891918</td>\n",
       "      <td>0.622772</td>\n",
       "      <td>0.808119</td>\n",
       "      <td>0.882790</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block cv</th>\n",
       "      <td>0.594799</td>\n",
       "      <td>0.793921</td>\n",
       "      <td>0.887450</td>\n",
       "      <td>0.580533</td>\n",
       "      <td>0.787131</td>\n",
       "      <td>0.873328</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>0.593454</td>\n",
       "      <td>0.793339</td>\n",
       "      <td>0.887681</td>\n",
       "      <td>0.563692</td>\n",
       "      <td>0.779500</td>\n",
       "      <td>0.874858</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>0.578570</td>\n",
       "      <td>0.785044</td>\n",
       "      <td>0.879517</td>\n",
       "      <td>0.507389</td>\n",
       "      <td>0.751345</td>\n",
       "      <td>0.861229</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   France                                  Spain            \\\n",
       "                      MCC        F1 Recall residential       MCC        F1   \n",
       "Model                                                                        \n",
       "Random cv        0.605411  0.799586           0.891918  0.622772  0.808119   \n",
       "Block cv         0.594799  0.793921           0.887450  0.580533  0.787131   \n",
       "Neighborhood cv  0.593454  0.793339           0.887681  0.563692  0.779500   \n",
       "City cv          0.578570  0.785044           0.879517  0.507389  0.751345   \n",
       "\n",
       "                                    \n",
       "                Recall residential  \n",
       "Model                               \n",
       "Random cv                 0.882790  \n",
       "Block cv                  0.873328  \n",
       "Neighborhood cv           0.874858  \n",
       "City cv                   0.861229  "
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type_class_results = pd.read_csv('../results/results-binary-type-country-comparison.csv')\n",
    "type_class_results['Country'] = type_class_results['name'].str.split('_').str[0]\n",
    "type_class_results['Model'] = type_class_results['name'].str.split('_').str[1]\n",
    "type_class_results['Model'] = type_class_results['Model'].str.replace('-', ' ').str.capitalize()\n",
    "type_class_results.rename(columns={'Recall_res': 'Recall residential'}, inplace=True)\n",
    "\n",
    "pivot_table = type_class_results.pivot_table(values=['MCC', 'F1', 'Recall residential'], index=['Model'], columns=['Country'], margins=False)\n",
    "pivot_table = pivot_table.swaplevel(0, 1, axis=1)\n",
    "pivot_table = pivot_table.sort_index(axis=1, level=0)\n",
    "\n",
    "pivot_table.columns = pivot_table.columns.rename(None, level=0) # drop Country cell\n",
    "\n",
    "# new_cols = pivot_table.columns.reindex(['France', 'Spain'])\n",
    "# pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "new_cols = pivot_table.columns.reindex(['MCC', 'F1', 'Recall residential'], level=1)\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "\n",
    "pivot_table = pivot_table.sort_values(by=('France', 'MCC'), ascending=False)\n",
    "pivot_table.to_latex('../figures/error-type-classification.tex', index=True, float_format='%.3f', caption='Binary type classification error', label='table:binary-type-classification-results')\n",
    "pivot_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 413,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In future versions `DataFrame.to_latex` is expected to utilise the base implementation of `Styler.to_latex` for formatting and rendering. The arguments signature may therefore change. It is recommended instead to use `DataFrame.style.to_latex` which also contains additional functionality.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>MCC</th>\n",
       "      <th>F1</th>\n",
       "      <th>Recall_residential</th>\n",
       "      <th>Recall_commercial</th>\n",
       "      <th>Recall_agricultural</th>\n",
       "      <th>Recall_industrial</th>\n",
       "      <th>Recall_others</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Country</th>\n",
       "      <th>Model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">France</th>\n",
       "      <th>Random cv</th>\n",
       "      <td>0.563697</td>\n",
       "      <td>0.571707</td>\n",
       "      <td>0.850827</td>\n",
       "      <td>0.598237</td>\n",
       "      <td>0.242210</td>\n",
       "      <td>0.612376</td>\n",
       "      <td>0.778976</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block cv</th>\n",
       "      <td>0.557402</td>\n",
       "      <td>0.523880</td>\n",
       "      <td>0.851716</td>\n",
       "      <td>0.554765</td>\n",
       "      <td>0.147083</td>\n",
       "      <td>0.472869</td>\n",
       "      <td>0.776849</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>0.555103</td>\n",
       "      <td>0.516221</td>\n",
       "      <td>0.852012</td>\n",
       "      <td>0.538304</td>\n",
       "      <td>0.146388</td>\n",
       "      <td>0.441206</td>\n",
       "      <td>0.775734</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>0.552792</td>\n",
       "      <td>0.503672</td>\n",
       "      <td>0.852605</td>\n",
       "      <td>0.531085</td>\n",
       "      <td>0.113394</td>\n",
       "      <td>0.409324</td>\n",
       "      <td>0.772660</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">Spain</th>\n",
       "      <th>Random cv</th>\n",
       "      <td>0.569145</td>\n",
       "      <td>0.649960</td>\n",
       "      <td>0.807103</td>\n",
       "      <td>0.583332</td>\n",
       "      <td>0.922713</td>\n",
       "      <td>0.704525</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Block cv</th>\n",
       "      <td>0.518983</td>\n",
       "      <td>0.599613</td>\n",
       "      <td>0.806647</td>\n",
       "      <td>0.428976</td>\n",
       "      <td>0.868593</td>\n",
       "      <td>0.630735</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Neighborhood cv</th>\n",
       "      <td>0.505610</td>\n",
       "      <td>0.590629</td>\n",
       "      <td>0.815331</td>\n",
       "      <td>0.399526</td>\n",
       "      <td>0.857954</td>\n",
       "      <td>0.583030</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>City cv</th>\n",
       "      <td>0.448355</td>\n",
       "      <td>0.548684</td>\n",
       "      <td>0.815391</td>\n",
       "      <td>0.338569</td>\n",
       "      <td>0.717360</td>\n",
       "      <td>0.542618</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                              MCC        F1  Recall_residential  \\\n",
       "Country Model                                                     \n",
       "France  Random cv        0.563697  0.571707            0.850827   \n",
       "        Block cv         0.557402  0.523880            0.851716   \n",
       "        Neighborhood cv  0.555103  0.516221            0.852012   \n",
       "        City cv          0.552792  0.503672            0.852605   \n",
       "Spain   Random cv        0.569145  0.649960            0.807103   \n",
       "        Block cv         0.518983  0.599613            0.806647   \n",
       "        Neighborhood cv  0.505610  0.590629            0.815331   \n",
       "        City cv          0.448355  0.548684            0.815391   \n",
       "\n",
       "                         Recall_commercial  Recall_agricultural  \\\n",
       "Country Model                                                     \n",
       "France  Random cv                 0.598237             0.242210   \n",
       "        Block cv                  0.554765             0.147083   \n",
       "        Neighborhood cv           0.538304             0.146388   \n",
       "        City cv                   0.531085             0.113394   \n",
       "Spain   Random cv                 0.583332             0.922713   \n",
       "        Block cv                  0.428976             0.868593   \n",
       "        Neighborhood cv           0.399526             0.857954   \n",
       "        City cv                   0.338569             0.717360   \n",
       "\n",
       "                         Recall_industrial  Recall_others  \n",
       "Country Model                                              \n",
       "France  Random cv                 0.612376       0.778976  \n",
       "        Block cv                  0.472869       0.776849  \n",
       "        Neighborhood cv           0.441206       0.775734  \n",
       "        City cv                   0.409324       0.772660  \n",
       "Spain   Random cv                 0.704525            NaN  \n",
       "        Block cv                  0.630735            NaN  \n",
       "        Neighborhood cv           0.583030            NaN  \n",
       "        City cv                   0.542618            NaN  "
      ]
     },
     "execution_count": 413,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type_class_results = pd.read_csv('../results/results-finegranular-type-country-comparison.csv')\n",
    "type_class_results['Country'] = type_class_results['name'].str.split('_').str[0]\n",
    "type_class_results['Model'] = type_class_results['name'].str.split('_').str[1]\n",
    "type_class_results['Model'] = type_class_results['Model'].str.replace('-', ' ').str.capitalize()\n",
    "type_class_results.rename(columns={'Recall_res': 'Recall residential'}, inplace=True)\n",
    "\n",
    "metrics = ['MCC', 'F1', 'Recall_residential', 'Recall_commercial', 'Recall_agricultural', 'Recall_industrial', 'Recall_others']\n",
    "pivot_table = type_class_results.pivot_table(values=metrics, index=['Country', 'Model'],  margins=False)\n",
    "pivot_table = pivot_table.sort_index(axis=1, level=0)\n",
    "\n",
    "# pivot_table.columns = pivot_table.index.rename(None, level=0) # drop Country cell\n",
    "\n",
    "# new_cols = pivot_table.columns.reindex(['France', 'Spain'])\n",
    "# pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "new_cols = pivot_table.columns.reindex(metrics)\n",
    "pivot_table = pivot_table.reindex(columns=new_cols[0])\n",
    "\n",
    "new_index = pivot_table.index.reindex(['Random cv', 'Block cv', 'Neighborhood cv', 'City cv'], level=1)\n",
    "pivot_table = pivot_table.reindex(new_index[0])\n",
    "\n",
    "pivot_table.to_latex('../figures/error-fine-granular-type-classification.tex', index=True, float_format='%.3f', caption='Fine-granular building type classification error', label='table:fine-granular-type-classification-results')\n",
    "pivot_table"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ------------------------------------------------------------------------------------------------------------------------"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Misc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Outlier Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 547,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "46035\n",
      "9139\n",
      "8631.0\n",
      "5.333680917622523\n",
      "566842\n",
      "7833\n",
      "14895.8\n",
      "38.05381382671626\n",
      "483856\n",
      "4133\n",
      "2909.0\n",
      "166.33069783430733\n"
     ]
    }
   ],
   "source": [
    "predictors = {'Netherlands': regressor_fts_importance_nl, 'France': regressor_fts_importance_fr, 'Spain': regressor_fts_importance_es}\n",
    "for c, p in predictors.items():\n",
    "    # inliers = remove_outliers(p.y_test).index\n",
    "    # # inliers = p.y_test[p.y_test['age'] < 2020].index\n",
    "    # outliers = p.y_test.index.difference(inliers)\n",
    "    # print(c)\n",
    "    # print('Overall MAE' , p.mae())\n",
    "    # print('Inliers MAE', metrics.mean_absolute_error(p.y_test.loc[inliers], p.y_predict.loc[inliers]))\n",
    "    # print('Outliers MAE', metrics.mean_absolute_error(p.y_test.loc[outliers], p.y_predict.loc[outliers]))\n",
    "    # print('Overall R2' , p.r2())\n",
    "    # print('Inliers R2', metrics.r2_score(p.y_test.loc[inliers], p.y_predict.loc[inliers]))\n",
    "    # print('Outliers R2', metrics.r2_score(p.y_test.loc[outliers], p.y_predict.loc[outliers]))\n",
    "    # print(p.r2())\n",
    "    # print(len(p.y_test[dataset.AGE_ATTRIBUTE][p.y_test[dataset.AGE_ATTRIBUTE] > 2020.0]))\n",
    "    n_1900 = len(p.y_test[dataset.AGE_ATTRIBUTE][p.y_test[dataset.AGE_ATTRIBUTE] == 1900.0])\n",
    "    print(n_1900)\n",
    "    n_1901 = len(p.y_test[dataset.AGE_ATTRIBUTE][p.y_test[dataset.AGE_ATTRIBUTE] == 1901.0])\n",
    "    print(n_1901)\n",
    "    n_190x = len(p.y_test[dataset.AGE_ATTRIBUTE][p.y_test[dataset.AGE_ATTRIBUTE].between(1900, 1905, inclusive='right')])/ 5\n",
    "    print(n_190x)\n",
    "    print(n_1900 / n_190x)\n",
    "    # print(len(p.y_test[dataset.AGE_ATTRIBUTE][p.y_test[dataset.AGE_ATTRIBUTE] == 1902.0]))\n",
    "    # print(len(p.y_test[dataset.AGE_ATTRIBUTE][p.y_test[dataset.AGE_ATTRIBUTE] == 1903.0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 464,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>error</th>\n",
       "      <th>y_test</th>\n",
       "      <th>y_predict</th>\n",
       "      <th>source_file</th>\n",
       "      <th>geometry</th>\n",
       "      <th>block</th>\n",
       "      <th>country</th>\n",
       "      <th>height</th>\n",
       "      <th>Unnamed: 0.1</th>\n",
       "      <th>sbb</th>\n",
       "      <th>id_source</th>\n",
       "      <th>city</th>\n",
       "      <th>id_misaligned</th>\n",
       "      <th>id</th>\n",
       "      <th>type_source</th>\n",
       "      <th>floors</th>\n",
       "      <th>TouchesIndexes</th>\n",
       "      <th>type</th>\n",
       "      <th>neighborhood</th>\n",
       "      <th>cv_fold_idx</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>v0.1-ESP.5.7.4.17_1-93</th>\n",
       "      <td>7224.803345</td>\n",
       "      <td>9183.0</td>\n",
       "      <td>1958.196655</td>\n",
       "      <td>A</td>\n",
       "      <td>POINT (3290315.567 2122459.722)</td>\n",
       "      <td>a28842a7</td>\n",
       "      <td>ESP</td>\n",
       "      <td>2.5</td>\n",
       "      <td>NaN</td>\n",
       "      <td>30263449</td>\n",
       "      <td>ES.SDGC.BU.4784706WL5848S</td>\n",
       "      <td>Momblona</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-93</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-93</td>\n",
       "      <td>3_industrial</td>\n",
       "      <td>1.0</td>\n",
       "      <td>[91, 92, 93]</td>\n",
       "      <td>non-residential</td>\n",
       "      <td>3faf6193</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-ESP.5.7.4.17_1-51</th>\n",
       "      <td>86.908325</td>\n",
       "      <td>1900.0</td>\n",
       "      <td>1986.908325</td>\n",
       "      <td>A</td>\n",
       "      <td>POINT (3290213.149 2122650.472)</td>\n",
       "      <td>f9245e50</td>\n",
       "      <td>ESP</td>\n",
       "      <td>2.5</td>\n",
       "      <td>NaN</td>\n",
       "      <td>a315b9e7</td>\n",
       "      <td>ES.SDGC.BU.4685901WL5848N</td>\n",
       "      <td>Momblona</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-51</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-51</td>\n",
       "      <td>3_industrial</td>\n",
       "      <td>1.0</td>\n",
       "      <td>[51, 52, 56, 57, 53, 54]</td>\n",
       "      <td>non-residential</td>\n",
       "      <td>3faf6193</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-ESP.5.7.4.17_1-41</th>\n",
       "      <td>85.884033</td>\n",
       "      <td>1900.0</td>\n",
       "      <td>1985.884033</td>\n",
       "      <td>A</td>\n",
       "      <td>POINT (3290199.294 2122645.096)</td>\n",
       "      <td>3a13a1ea</td>\n",
       "      <td>ESP</td>\n",
       "      <td>5.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>48cfe5ec</td>\n",
       "      <td>ES.SDGC.BU.4685201WL5848N</td>\n",
       "      <td>Momblona</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-41</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-41</td>\n",
       "      <td>3_industrial</td>\n",
       "      <td>2.0</td>\n",
       "      <td>[41, 43, 48]</td>\n",
       "      <td>non-residential</td>\n",
       "      <td>3faf6193</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-ESP.5.7.4.17_1-55</th>\n",
       "      <td>70.786255</td>\n",
       "      <td>1929.0</td>\n",
       "      <td>1999.786255</td>\n",
       "      <td>A</td>\n",
       "      <td>POINT (3290210.422 2122621.881)</td>\n",
       "      <td>aa0be5e9</td>\n",
       "      <td>ESP</td>\n",
       "      <td>2.5</td>\n",
       "      <td>NaN</td>\n",
       "      <td>a315b9e7</td>\n",
       "      <td>ES.SDGC.BU.4685905WL5848N</td>\n",
       "      <td>Momblona</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-55</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-55</td>\n",
       "      <td>1_residential</td>\n",
       "      <td>1.0</td>\n",
       "      <td>[55, 66]</td>\n",
       "      <td>residential</td>\n",
       "      <td>3faf6193</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v0.1-ESP.5.7.4.17_1-59</th>\n",
       "      <td>63.993774</td>\n",
       "      <td>1900.0</td>\n",
       "      <td>1963.993774</td>\n",
       "      <td>A</td>\n",
       "      <td>POINT (3290250.649 2122559.782)</td>\n",
       "      <td>93c7264c</td>\n",
       "      <td>ESP</td>\n",
       "      <td>5.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>a315b9e7</td>\n",
       "      <td>ES.SDGC.BU.4685908WL5848S</td>\n",
       "      <td>Momblona</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-59</td>\n",
       "      <td>v0.1-ESP.5.7.4.17_1-59</td>\n",
       "      <td>1_residential</td>\n",
       "      <td>2.0</td>\n",
       "      <td>[59]</td>\n",
       "      <td>residential</td>\n",
       "      <td>3faf6193</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                              error  y_test    y_predict source_file  \\\n",
       "id                                                                     \n",
       "v0.1-ESP.5.7.4.17_1-93  7224.803345  9183.0  1958.196655           A   \n",
       "v0.1-ESP.5.7.4.17_1-51    86.908325  1900.0  1986.908325           A   \n",
       "v0.1-ESP.5.7.4.17_1-41    85.884033  1900.0  1985.884033           A   \n",
       "v0.1-ESP.5.7.4.17_1-55    70.786255  1929.0  1999.786255           A   \n",
       "v0.1-ESP.5.7.4.17_1-59    63.993774  1900.0  1963.993774           A   \n",
       "\n",
       "                                               geometry     block country  \\\n",
       "id                                                                          \n",
       "v0.1-ESP.5.7.4.17_1-93  POINT (3290315.567 2122459.722)  a28842a7     ESP   \n",
       "v0.1-ESP.5.7.4.17_1-51  POINT (3290213.149 2122650.472)  f9245e50     ESP   \n",
       "v0.1-ESP.5.7.4.17_1-41  POINT (3290199.294 2122645.096)  3a13a1ea     ESP   \n",
       "v0.1-ESP.5.7.4.17_1-55  POINT (3290210.422 2122621.881)  aa0be5e9     ESP   \n",
       "v0.1-ESP.5.7.4.17_1-59  POINT (3290250.649 2122559.782)  93c7264c     ESP   \n",
       "\n",
       "                        height  Unnamed: 0.1       sbb  \\\n",
       "id                                                       \n",
       "v0.1-ESP.5.7.4.17_1-93     2.5           NaN  30263449   \n",
       "v0.1-ESP.5.7.4.17_1-51     2.5           NaN  a315b9e7   \n",
       "v0.1-ESP.5.7.4.17_1-41     5.0           NaN  48cfe5ec   \n",
       "v0.1-ESP.5.7.4.17_1-55     2.5           NaN  a315b9e7   \n",
       "v0.1-ESP.5.7.4.17_1-59     5.0           NaN  a315b9e7   \n",
       "\n",
       "                                        id_source      city  \\\n",
       "id                                                            \n",
       "v0.1-ESP.5.7.4.17_1-93  ES.SDGC.BU.4784706WL5848S  Momblona   \n",
       "v0.1-ESP.5.7.4.17_1-51  ES.SDGC.BU.4685901WL5848N  Momblona   \n",
       "v0.1-ESP.5.7.4.17_1-41  ES.SDGC.BU.4685201WL5848N  Momblona   \n",
       "v0.1-ESP.5.7.4.17_1-55  ES.SDGC.BU.4685905WL5848N  Momblona   \n",
       "v0.1-ESP.5.7.4.17_1-59  ES.SDGC.BU.4685908WL5848S  Momblona   \n",
       "\n",
       "                                 id_misaligned                      id  \\\n",
       "id                                                                       \n",
       "v0.1-ESP.5.7.4.17_1-93  v0.1-ESP.5.7.4.17_1-93  v0.1-ESP.5.7.4.17_1-93   \n",
       "v0.1-ESP.5.7.4.17_1-51  v0.1-ESP.5.7.4.17_1-51  v0.1-ESP.5.7.4.17_1-51   \n",
       "v0.1-ESP.5.7.4.17_1-41  v0.1-ESP.5.7.4.17_1-41  v0.1-ESP.5.7.4.17_1-41   \n",
       "v0.1-ESP.5.7.4.17_1-55  v0.1-ESP.5.7.4.17_1-55  v0.1-ESP.5.7.4.17_1-55   \n",
       "v0.1-ESP.5.7.4.17_1-59  v0.1-ESP.5.7.4.17_1-59  v0.1-ESP.5.7.4.17_1-59   \n",
       "\n",
       "                          type_source  floors            TouchesIndexes  \\\n",
       "id                                                                        \n",
       "v0.1-ESP.5.7.4.17_1-93   3_industrial     1.0              [91, 92, 93]   \n",
       "v0.1-ESP.5.7.4.17_1-51   3_industrial     1.0  [51, 52, 56, 57, 53, 54]   \n",
       "v0.1-ESP.5.7.4.17_1-41   3_industrial     2.0              [41, 43, 48]   \n",
       "v0.1-ESP.5.7.4.17_1-55  1_residential     1.0                  [55, 66]   \n",
       "v0.1-ESP.5.7.4.17_1-59  1_residential     2.0                      [59]   \n",
       "\n",
       "                                   type neighborhood  cv_fold_idx  \n",
       "id                                                                 \n",
       "v0.1-ESP.5.7.4.17_1-93  non-residential     3faf6193            2  \n",
       "v0.1-ESP.5.7.4.17_1-51  non-residential     3faf6193            2  \n",
       "v0.1-ESP.5.7.4.17_1-41  non-residential     3faf6193            2  \n",
       "v0.1-ESP.5.7.4.17_1-55      residential     3faf6193            2  \n",
       "v0.1-ESP.5.7.4.17_1-59      residential     3faf6193            2  "
      ]
     },
     "execution_count": 464,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# error_df[error_df['Country'] == 'Spain'].sort_values(by='MAE', ascending=False).head(20)\n",
    "df = regressor_fts_importance_es.individual_prediction_error()\n",
    "df['y_test'] = regressor_fts_importance_es.y_test[dataset.AGE_ATTRIBUTE]\n",
    "df['y_predict'] = regressor_fts_importance_es.y_predict[dataset.AGE_ATTRIBUTE]\n",
    "df = df.join(regressor_fts_importance_es.aux_vars_test, how=\"inner\")\n",
    "df['error'] = df['error'].abs()\n",
    "df[df['city'] == 'Momblona'].sort_values(by='error', ascending=False).head(5)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ID issues"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'df_nl_old' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb Cell 49'\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000032?line=0'>1</a>\u001b[0m df_nl_old[\u001b[39m'\u001b[39m\u001b[39mid\u001b[39m\u001b[39m'\u001b[39m] \u001b[39m=\u001b[39m df_nl_old[\u001b[39m'\u001b[39m\u001b[39mid\u001b[39m\u001b[39m'\u001b[39m]\u001b[39m.\u001b[39mastype(\u001b[39mstr\u001b[39m)\n\u001b[1;32m      <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000032?line=1'>2</a>\u001b[0m old \u001b[39m=\u001b[39m df_nl_old[df_nl_old[\u001b[39m'\u001b[39m\u001b[39mid\u001b[39m\u001b[39m'\u001b[39m] \u001b[39m==\u001b[39m \u001b[39m'\u001b[39m\u001b[39m367425\u001b[39m\u001b[39m'\u001b[39m]\u001b[39m.\u001b[39mreset_index(drop\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m)\u001b[39m.\u001b[39m_get_numeric_data()\n\u001b[1;32m      <a href='vscode-notebook-cell:/Users/D062794/workspace/ufo-prediction/notebooks/v0_1_prediction.ipynb#ch0000032?line=2'>3</a>\u001b[0m new \u001b[39m=\u001b[39m df_nl_new[df_nl_new[\u001b[39m'\u001b[39m\u001b[39mid\u001b[39m\u001b[39m'\u001b[39m] \u001b[39m==\u001b[39m \u001b[39m'\u001b[39m\u001b[39mv0.1-NLD.4.2_1-0\u001b[39m\u001b[39m'\u001b[39m]\u001b[39m.\u001b[39mreset_index(drop\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m)\u001b[39m.\u001b[39m_get_numeric_data()\n",
      "\u001b[0;31mNameError\u001b[0m: name 'df_nl_old' is not defined"
     ]
    }
   ],
   "source": [
    "df_nl_old['id'] = df_nl_old['id'].astype(str)\n",
    "old = df_nl_old[df_nl_old['id'] == '367425'].reset_index(drop=True)._get_numeric_data()\n",
    "new = df_nl_new[df_nl_new['id'] == 'v0.1-NLD.4.2_1-0'].reset_index(drop=True)._get_numeric_data()\n",
    "old.subtract(new).div(old).multiply(100).abs().sort_values(ascending=False, by=0, axis=1)\n",
    "\n",
    "# -> features seem to consistently have new ID, only Angerlo_attrib_source.csv and Angerlo_attrib.csv seems to be messed up (and maybe buffer.csv)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>id_source</th>\n",
       "      <th>FootprintArea</th>\n",
       "      <th>buildings_within_buffer_100</th>\n",
       "      <th>total_ft_area_within_buffer_100</th>\n",
       "      <th>std_elongation_within_buffer_500</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>v0.1-NLD.4.2_1-0</td>\n",
       "      <td>362276</td>\n",
       "      <td>67.830371</td>\n",
       "      <td>174.0</td>\n",
       "      <td>9112.819401</td>\n",
       "      <td>0.184137</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>v0.1-NLD.4.2_1-1</td>\n",
       "      <td>368961</td>\n",
       "      <td>154.564253</td>\n",
       "      <td>7.0</td>\n",
       "      <td>760.960856</td>\n",
       "      <td>0.192126</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>v0.1-NLD.4.2_1-2</td>\n",
       "      <td>11274732</td>\n",
       "      <td>808.611675</td>\n",
       "      <td>5.0</td>\n",
       "      <td>3898.621112</td>\n",
       "      <td>0.180159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>v0.1-NLD.4.2_1-3</td>\n",
       "      <td>11274834</td>\n",
       "      <td>107.518723</td>\n",
       "      <td>164.0</td>\n",
       "      <td>8072.889501</td>\n",
       "      <td>0.189038</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>v0.1-NLD.4.2_1-4</td>\n",
       "      <td>7087934</td>\n",
       "      <td>116.648825</td>\n",
       "      <td>47.0</td>\n",
       "      <td>4620.357978</td>\n",
       "      <td>0.181643</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4032</th>\n",
       "      <td>v0.1-NLD.4.2_1-4032</td>\n",
       "      <td>11278003</td>\n",
       "      <td>102.994828</td>\n",
       "      <td>14.0</td>\n",
       "      <td>1267.060666</td>\n",
       "      <td>0.201105</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4033</th>\n",
       "      <td>v0.1-NLD.4.2_1-4033</td>\n",
       "      <td>17389836</td>\n",
       "      <td>90.836315</td>\n",
       "      <td>11.0</td>\n",
       "      <td>4174.980928</td>\n",
       "      <td>0.169502</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4034</th>\n",
       "      <td>v0.1-NLD.4.2_1-4034</td>\n",
       "      <td>11278039</td>\n",
       "      <td>69.628687</td>\n",
       "      <td>133.0</td>\n",
       "      <td>5778.010517</td>\n",
       "      <td>0.185368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4035</th>\n",
       "      <td>v0.1-NLD.4.2_1-4035</td>\n",
       "      <td>17381816</td>\n",
       "      <td>10.593620</td>\n",
       "      <td>83.0</td>\n",
       "      <td>4952.143570</td>\n",
       "      <td>0.170931</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4036</th>\n",
       "      <td>v0.1-NLD.4.2_1-4036</td>\n",
       "      <td>366773</td>\n",
       "      <td>73.006670</td>\n",
       "      <td>106.0</td>\n",
       "      <td>8154.858035</td>\n",
       "      <td>0.164236</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>4037 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                       id id_source  FootprintArea  \\\n",
       "0        v0.1-NLD.4.2_1-0    362276      67.830371   \n",
       "1        v0.1-NLD.4.2_1-1    368961     154.564253   \n",
       "2        v0.1-NLD.4.2_1-2  11274732     808.611675   \n",
       "3        v0.1-NLD.4.2_1-3  11274834     107.518723   \n",
       "4        v0.1-NLD.4.2_1-4   7087934     116.648825   \n",
       "...                   ...       ...            ...   \n",
       "4032  v0.1-NLD.4.2_1-4032  11278003     102.994828   \n",
       "4033  v0.1-NLD.4.2_1-4033  17389836      90.836315   \n",
       "4034  v0.1-NLD.4.2_1-4034  11278039      69.628687   \n",
       "4035  v0.1-NLD.4.2_1-4035  17381816      10.593620   \n",
       "4036  v0.1-NLD.4.2_1-4036    366773      73.006670   \n",
       "\n",
       "      buildings_within_buffer_100  total_ft_area_within_buffer_100  \\\n",
       "0                           174.0                      9112.819401   \n",
       "1                             7.0                       760.960856   \n",
       "2                             5.0                      3898.621112   \n",
       "3                           164.0                      8072.889501   \n",
       "4                            47.0                      4620.357978   \n",
       "...                           ...                              ...   \n",
       "4032                         14.0                      1267.060666   \n",
       "4033                         11.0                      4174.980928   \n",
       "4034                        133.0                      5778.010517   \n",
       "4035                         83.0                      4952.143570   \n",
       "4036                        106.0                      8154.858035   \n",
       "\n",
       "      std_elongation_within_buffer_500  \n",
       "0                             0.184137  \n",
       "1                             0.192126  \n",
       "2                             0.180159  \n",
       "3                             0.189038  \n",
       "4                             0.181643  \n",
       "...                                ...  \n",
       "4032                          0.201105  \n",
       "4033                          0.169502  \n",
       "4034                          0.185368  \n",
       "4035                          0.170931  \n",
       "4036                          0.164236  \n",
       "\n",
       "[4037 rows x 6 columns]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = data_netherlands\n",
    "# df[df['id_source'].isin([363672, 17385592, 367425, 15587580, 17390370])]\n",
    "df[['id', 'id_source', 'FootprintArea', 'buildings_within_buffer_100', 'total_ft_area_within_buffer_100', 'std_elongation_within_buffer_500']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-11 09:56:36,668 | INFO : Starting experiment 1-cities_...\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:36,882 | INFO : Dataset length: 4437\n",
      "2022-07-11 09:56:36,889 | INFO : Dataset allocated memory: 4 MB\n",
      "2022-07-11 09:56:36,892 | INFO : Dataset standard deviation: 30.566045986137226\n",
      "2022-07-11 09:56:36,894 | INFO : Dataset mean age: 1966.5158891142664\n",
      "2022-07-11 09:56:36,895 | INFO : Training dataset length: 4437\n",
      "2022-07-11 09:56:36,896 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:37,387 | INFO : Test dataset standard deviation after preprocessing: 27.750464681809287\n",
      "2022-07-11 09:56:37,388 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:37,389 | INFO : Training dataset length after preprocessing: 4352\n",
      "2022-07-11 09:56:37,390 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:39,659 | INFO : Dataset length: 10643\n",
      "2022-07-11 09:56:39,664 | INFO : Dataset allocated memory: 11 MB\n",
      "2022-07-11 09:56:39,667 | INFO : Dataset standard deviation: 25.28004933771152\n",
      "2022-07-11 09:56:39,667 | INFO : Dataset mean age: 1977.3629615709856\n",
      "2022-07-11 09:56:39,668 | INFO : Training dataset length: 10643\n",
      "2022-07-11 09:56:39,668 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:39,951 | INFO : Test dataset standard deviation after preprocessing: 27.750464681809287\n",
      "2022-07-11 09:56:39,952 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:39,953 | INFO : Training dataset length after preprocessing: 10595\n",
      "2022-07-11 09:56:39,954 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:41,637 | INFO : Dataset length: 3610\n",
      "2022-07-11 09:56:41,641 | INFO : Dataset allocated memory: 3 MB\n",
      "2022-07-11 09:56:41,643 | INFO : Dataset standard deviation: 23.413806835798336\n",
      "2022-07-11 09:56:41,643 | INFO : Dataset mean age: 1969.4188365650969\n",
      "2022-07-11 09:56:41,644 | INFO : Training dataset length: 3610\n",
      "2022-07-11 09:56:41,644 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:41,931 | INFO : Test dataset standard deviation after preprocessing: 27.75046468180929\n",
      "2022-07-11 09:56:41,932 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:41,933 | INFO : Training dataset length after preprocessing: 3595\n",
      "2022-07-11 09:56:41,934 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-11 09:56:43,998 | INFO : Starting experiment 2-cities_...\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:44,142 | INFO : Dataset length: 9979\n",
      "2022-07-11 09:56:44,148 | INFO : Dataset allocated memory: 10 MB\n",
      "2022-07-11 09:56:44,149 | INFO : Dataset standard deviation: 32.25948298605961\n",
      "2022-07-11 09:56:44,150 | INFO : Dataset mean age: 1969.2166549754484\n",
      "2022-07-11 09:56:44,151 | INFO : Training dataset length: 9979\n",
      "2022-07-11 09:56:44,152 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:44,553 | INFO : Test dataset standard deviation after preprocessing: 27.750464681809287\n",
      "2022-07-11 09:56:44,556 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:44,556 | INFO : Training dataset length after preprocessing: 9784\n",
      "2022-07-11 09:56:44,661 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:46,549 | INFO : Dataset length: 32729\n",
      "2022-07-11 09:56:46,555 | INFO : Dataset allocated memory: 34 MB\n",
      "2022-07-11 09:56:46,556 | INFO : Dataset standard deviation: 19.495191960699348\n",
      "2022-07-11 09:56:46,557 | INFO : Dataset mean age: 1983.5036511961869\n",
      "2022-07-11 09:56:46,558 | INFO : Training dataset length: 32729\n",
      "2022-07-11 09:56:46,558 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:46,881 | INFO : Test dataset standard deviation after preprocessing: 27.750464681809287\n",
      "2022-07-11 09:56:46,882 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:46,883 | INFO : Training dataset length after preprocessing: 32681\n",
      "2022-07-11 09:56:46,883 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:48,784 | INFO : Dataset length: 25696\n",
      "2022-07-11 09:56:48,789 | INFO : Dataset allocated memory: 26 MB\n",
      "2022-07-11 09:56:48,792 | INFO : Dataset standard deviation: 17.559509510141613\n",
      "2022-07-11 09:56:48,793 | INFO : Dataset mean age: 1984.0682985678704\n",
      "2022-07-11 09:56:48,793 | INFO : Training dataset length: 25696\n",
      "2022-07-11 09:56:48,794 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:49,094 | INFO : Test dataset standard deviation after preprocessing: 27.75046468180929\n",
      "2022-07-11 09:56:49,096 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:49,097 | INFO : Training dataset length after preprocessing: 25681\n",
      "2022-07-11 09:56:49,097 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-11 09:56:51,386 | INFO : Starting experiment 4-cities_...\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:51,532 | INFO : Dataset length: 24232\n",
      "2022-07-11 09:56:51,538 | INFO : Dataset allocated memory: 25 MB\n",
      "2022-07-11 09:56:51,540 | INFO : Dataset standard deviation: 28.407737907087906\n",
      "2022-07-11 09:56:51,540 | INFO : Dataset mean age: 1972.8247358864312\n",
      "2022-07-11 09:56:51,541 | INFO : Training dataset length: 24232\n",
      "2022-07-11 09:56:51,542 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:51,835 | INFO : Test dataset standard deviation after preprocessing: 27.750464681809287\n",
      "2022-07-11 09:56:51,836 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:51,837 | INFO : Training dataset length after preprocessing: 23974\n",
      "2022-07-11 09:56:51,838 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:53,806 | INFO : Dataset length: 41881\n",
      "2022-07-11 09:56:53,812 | INFO : Dataset allocated memory: 43 MB\n",
      "2022-07-11 09:56:53,814 | INFO : Dataset standard deviation: 22.813369325698563\n",
      "2022-07-11 09:56:53,814 | INFO : Dataset mean age: 1980.6851555598005\n",
      "2022-07-11 09:56:53,815 | INFO : Training dataset length: 41881\n",
      "2022-07-11 09:56:53,816 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:54,162 | INFO : Test dataset standard deviation after preprocessing: 27.750464681809287\n",
      "2022-07-11 09:56:54,163 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:54,164 | INFO : Training dataset length after preprocessing: 41708\n",
      "2022-07-11 09:56:54,165 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:56,190 | INFO : Dataset length: 41881\n",
      "2022-07-11 09:56:56,195 | INFO : Dataset allocated memory: 43 MB\n",
      "2022-07-11 09:56:56,198 | INFO : Dataset standard deviation: 22.813369325698563\n",
      "2022-07-11 09:56:56,200 | INFO : Dataset mean age: 1980.6851555598005\n",
      "2022-07-11 09:56:56,200 | INFO : Training dataset length: 41881\n",
      "2022-07-11 09:56:56,201 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:56,537 | INFO : Test dataset standard deviation after preprocessing: 27.75046468180929\n",
      "2022-07-11 09:56:56,539 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:56,539 | INFO : Training dataset length after preprocessing: 41708\n",
      "2022-07-11 09:56:56,540 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2022-07-11 09:56:58,940 | INFO : Starting experiment 8-cities_...\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:56:59,162 | INFO : Dataset length: 67855\n",
      "2022-07-11 09:56:59,169 | INFO : Dataset allocated memory: 70 MB\n",
      "2022-07-11 09:56:59,173 | INFO : Dataset standard deviation: 26.199331378912724\n",
      "2022-07-11 09:56:59,175 | INFO : Dataset mean age: 1979.2070002210596\n",
      "2022-07-11 09:56:59,176 | INFO : Training dataset length: 67855\n",
      "2022-07-11 09:56:59,176 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:56:59,560 | INFO : Test dataset standard deviation after preprocessing: 27.750464681809287\n",
      "2022-07-11 09:56:59,562 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:56:59,563 | INFO : Training dataset length after preprocessing: 67275\n",
      "2022-07-11 09:56:59,563 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:57:01,962 | INFO : Dataset length: 71736\n",
      "2022-07-11 09:57:01,967 | INFO : Dataset allocated memory: 74 MB\n",
      "2022-07-11 09:57:01,971 | INFO : Dataset standard deviation: 26.45701823864484\n",
      "2022-07-11 09:57:01,972 | INFO : Dataset mean age: 1978.4471395115422\n",
      "2022-07-11 09:57:01,972 | INFO : Training dataset length: 71736\n",
      "2022-07-11 09:57:01,973 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:57:02,387 | INFO : Test dataset standard deviation after preprocessing: 27.750464681809287\n",
      "2022-07-11 09:57:02,389 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:57:02,390 | INFO : Training dataset length after preprocessing: 71142\n",
      "2022-07-11 09:57:02,391 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "\n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "2022-07-11 09:57:05,057 | INFO : Dataset length: 58690\n",
      "2022-07-11 09:57:05,063 | INFO : Dataset allocated memory: 61 MB\n",
      "2022-07-11 09:57:05,065 | INFO : Dataset standard deviation: 25.663402602171793\n",
      "2022-07-11 09:57:05,066 | INFO : Dataset mean age: 1977.7439427500426\n",
      "2022-07-11 09:57:05,067 | INFO : Training dataset length: 58690\n",
      "2022-07-11 09:57:05,067 | INFO : Test dataset length: 102134\n",
      "2022-07-11 09:57:05,446 | INFO : Test dataset standard deviation after preprocessing: 27.75046468180929\n",
      "2022-07-11 09:57:05,447 | INFO : Test dataset mean age after preprocessing: 1971.6868309561492\n",
      "2022-07-11 09:57:05,448 | INFO : Training dataset length after preprocessing: 58239\n",
      "2022-07-11 09:57:05,449 | INFO : Test dataset length after preprocessing: 101731\n",
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n"
     ]
    }
   ],
   "source": [
    "EXPERIMENT = 'generalization'\n",
    "# df_test_set = utils.sample_cities(df_fr_sample, n=1)\n",
    "df_test_set = utils.sample_cities(df_nl_sample, n=1)\n",
    "\n",
    "comparison_config = {}\n",
    "for i in range(4):\n",
    "    comparison_config[f'{2**i}-cities'] = {'n_cities': 2**i}\n",
    "\n",
    "comparison = AgePredictorComparison(\n",
    "    model=XGBRegressor(tree_method='hist'),\n",
    "    df=df_nl_sample[~df_nl_sample['city'].isin(df_test_set.city.unique())],\n",
    "    n_seeds=3,\n",
    "    cross_validation_split=None if 'generalization' in EXPERIMENT else cross_validation,\n",
    "    test_set=df_test_set if 'generalization' in EXPERIMENT else None,\n",
    "    preprocessing_stages=[remove_buildings_pre_1900],\n",
    "    include_baseline=False,\n",
    "    compare_error_distribution=True,\n",
    "    garbage_collect_after_training=True,\n",
    "    comparison_config=comparison_config,\n",
    ")\n",
    "\n",
    "# comparison = AgeClassifierComparison(\n",
    "#     model=XGBClassifier(tree_method='hist'),\n",
    "#     df=df_nl_sample,\n",
    "#     bin_config=(1900, 2020, 50),\n",
    "#     n_seeds=3,\n",
    "#     cross_validation_split=cross_validation,\n",
    "#     preprocessing_stages=[remove_buildings_pre_1900],\n",
    "#     mitigate_class_imbalance=True,\n",
    "#     include_baseline=False,\n",
    "#     comparison_config=comparison_config,\n",
    "# )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>R2</th>\n",
       "      <th>R2_std</th>\n",
       "      <th>MAE</th>\n",
       "      <th>MAE_std</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>RMSE_std</th>\n",
       "      <th>within_5_years</th>\n",
       "      <th>within_10_years</th>\n",
       "      <th>within_20_years</th>\n",
       "      <th>R2_seed_0</th>\n",
       "      <th>R2_seed_1</th>\n",
       "      <th>R2_seed_2</th>\n",
       "      <th>skew</th>\n",
       "      <th>kurtosis</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1-cities_</td>\n",
       "      <td>-0.368240</td>\n",
       "      <td>0.263620</td>\n",
       "      <td>26.263298</td>\n",
       "      <td>3.175047</td>\n",
       "      <td>32.304229</td>\n",
       "      <td>3.176793</td>\n",
       "      <td>0.125969</td>\n",
       "      <td>0.241798</td>\n",
       "      <td>0.440944</td>\n",
       "      <td>-0.675454</td>\n",
       "      <td>-0.031718</td>\n",
       "      <td>-0.397548</td>\n",
       "      <td>-0.270973</td>\n",
       "      <td>-0.033941</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2-cities_</td>\n",
       "      <td>-0.111284</td>\n",
       "      <td>0.135737</td>\n",
       "      <td>22.801350</td>\n",
       "      <td>1.657933</td>\n",
       "      <td>29.200930</td>\n",
       "      <td>1.756155</td>\n",
       "      <td>0.175921</td>\n",
       "      <td>0.318566</td>\n",
       "      <td>0.528167</td>\n",
       "      <td>-0.301321</td>\n",
       "      <td>0.007214</td>\n",
       "      <td>-0.039745</td>\n",
       "      <td>-0.393378</td>\n",
       "      <td>-0.072832</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4-cities_</td>\n",
       "      <td>0.047493</td>\n",
       "      <td>0.013341</td>\n",
       "      <td>21.367119</td>\n",
       "      <td>0.244922</td>\n",
       "      <td>27.082666</td>\n",
       "      <td>0.190120</td>\n",
       "      <td>0.176164</td>\n",
       "      <td>0.330676</td>\n",
       "      <td>0.550986</td>\n",
       "      <td>0.036380</td>\n",
       "      <td>0.039847</td>\n",
       "      <td>0.066254</td>\n",
       "      <td>-0.329033</td>\n",
       "      <td>-0.237558</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>8-cities_</td>\n",
       "      <td>0.066072</td>\n",
       "      <td>0.017704</td>\n",
       "      <td>20.801898</td>\n",
       "      <td>0.218343</td>\n",
       "      <td>26.816709</td>\n",
       "      <td>0.253500</td>\n",
       "      <td>0.182314</td>\n",
       "      <td>0.343445</td>\n",
       "      <td>0.572572</td>\n",
       "      <td>0.074147</td>\n",
       "      <td>0.082559</td>\n",
       "      <td>0.041509</td>\n",
       "      <td>-0.474705</td>\n",
       "      <td>0.017343</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        name        R2    R2_std        MAE   MAE_std       RMSE  RMSE_std  \\\n",
       "0  1-cities_ -0.368240  0.263620  26.263298  3.175047  32.304229  3.176793   \n",
       "1  2-cities_ -0.111284  0.135737  22.801350  1.657933  29.200930  1.756155   \n",
       "2  4-cities_  0.047493  0.013341  21.367119  0.244922  27.082666  0.190120   \n",
       "3  8-cities_  0.066072  0.017704  20.801898  0.218343  26.816709  0.253500   \n",
       "\n",
       "   within_5_years  within_10_years  within_20_years  R2_seed_0  R2_seed_1  \\\n",
       "0        0.125969         0.241798         0.440944  -0.675454  -0.031718   \n",
       "1        0.175921         0.318566         0.528167  -0.301321   0.007214   \n",
       "2        0.176164         0.330676         0.550986   0.036380   0.039847   \n",
       "3        0.182314         0.343445         0.572572   0.074147   0.082559   \n",
       "\n",
       "   R2_seed_2      skew  kurtosis  \n",
       "0  -0.397548 -0.270973 -0.033941  \n",
       "1  -0.039745 -0.393378 -0.072832  \n",
       "2   0.066254 -0.329033 -0.237558  \n",
       "3   0.041509 -0.474705  0.017343  "
      ]
     },
     "execution_count": 160,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "comparison.evaluate()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.8.12 ('uf-ml')",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.12"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "d0d7e3373a39644d79cefd3d7fbe538923db789a925e1acd6f0b7d5044eb032b"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}