Branched Keras.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pickle\n",
    "import numpy as np\n",
    "np.random.seed(42)\n",
    "import pandas as pd\n",
    "import keras\n",
    "import seaborn as sn\n",
    "from keras.models import Model, Sequential\n",
    "from keras.layers import Embedding, Dense, Dropout, Input, Concatenate, Flatten\n",
    "from keras.callbacks import EarlyStopping\n",
    "from keras.utils import plot_model\n",
    "from sklearn.model_selection import train_test_split as splt\n",
    "from sklearn.impute import SimpleImputer\n",
    "import matplotlib.pyplot as plt\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"2\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "dfx = pd.read_excel(\"Data/Milano_Cleaned_wSCORE.xlsx\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Thresholds are 0 - 0.01, 0.01 - 0.02, 0.02 - 0.05, 0.05 - 1\n",
    "y = dfx[\"SCORE\"].values/100\n",
    "y_cat = np.zeros((y.shape[0], 4))\n",
    "for i, val in enumerate(y):\n",
    "    if(val < 0.01):\n",
    "        y_cat[i][0] = 1\n",
    "    elif(val < 0.02):\n",
    "        y_cat[i][1] = 1\n",
    "    elif(val < 0.05):\n",
    "        y_cat[i][2] = 1\n",
    "    else:\n",
    "        y_cat[i][3] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 5780 entries, 0 to 5779\n",
      "Data columns (total 63 columns):\n",
      "subject_id                        5780 non-null int64\n",
      "fibers_a                          5780 non-null int64\n",
      "NS_Drugs                          5780 non-null int64\n",
      "IMA                               5780 non-null int64\n",
      "HDL                               5780 non-null int64\n",
      "silent_ischaemia                  5780 non-null int64\n",
      "hepatopathies_YN                  5780 non-null int64\n",
      "remote_bypass_to                  5780 non-null int64\n",
      "Remote_PTCA_to                    5780 non-null int64\n",
      "extremities                       5780 non-null int64\n",
      "Remote_TIA_to                     5780 non-null int64\n",
      "age                               5780 non-null int64\n",
      "remote_peripheral_arteriopathy    5780 non-null int64\n",
      "antihypertensive_to               5780 non-null int64\n",
      "COPD                              5780 non-null int64\n",
      "Profile_bypass                    5780 non-null int64\n",
      "respiratory_system                5780 non-null int64\n",
      "statins_a                         5780 non-null int64\n",
      "right_corneal_arch                5780 non-null int64\n",
      "TIA_profile                       5780 non-null int64\n",
      "angina                            5780 non-null int64\n",
      "self_obesity                      5780 non-null int64\n",
      "thyroid_nodules                   5780 non-null int64\n",
      "hypertension_self                 5780 non-null int64\n",
      "diabetes2_self                    5780 non-null int64\n",
      "ischaemia_silent_profile          5780 non-null int64\n",
      "total_cholesterol                 5780 non-null int64\n",
      "TG                                5780 non-null int64\n",
      "aortic_arteriopathy               5780 non-null int64\n",
      "visits_carried_out                5780 non-null int64\n",
      "IMA_profile                       5780 non-null int64\n",
      "strokes                           5780 non-null int64\n",
      "cerebral_hemorrhage               5780 non-null int64\n",
      "standardised_visits_to            5780 non-null int64\n",
      "PTCA                              5780 non-null int64\n",
      "antiaggregants_to                 5780 non-null int64\n",
      "atrial_fibrillation               5780 non-null int64\n",
      "angina_profile                    5780 non-null int64\n",
      "CHD_profile                       5780 non-null int64\n",
      "Remote_stroke_at                  5780 non-null int64\n",
      "thyroid_function                  5780 non-null int64\n",
      "imbalance                         5780 non-null int64\n",
      "thyroid_appearance                5780 non-null int64\n",
      "peripheral_arterial_wrists        5780 non-null int64\n",
      "bypass                            5780 non-null int64\n",
      "antidiabetics_to                  5780 non-null int64\n",
      "xanthomatosis                     5780 non-null int64\n",
      "left_corneal_arch                 5780 non-null int64\n",
      "gender                            5780 non-null int64\n",
      "TIA                               5780 non-null int64\n",
      "CHD                               5780 non-null int64\n",
      "cholecystectomy                   5780 non-null int64\n",
      "remote_decompensation_at          5780 non-null int64\n",
      "Stroke_Profile                    5780 non-null int64\n",
      "remote_aortic_arteriopathy        5780 non-null int64\n",
      "education                         5780 non-null int64\n",
      "IVS                               5780 non-null int64\n",
      "remote_angina_a                   5780 non-null int64\n",
      "Remote_CHD_to                     5780 non-null int64\n",
      "steatosis                         5780 non-null int64\n",
      "remote_silent_ischemia_a          5780 non-null int64\n",
      "nephropathies_YN                  5780 non-null int64\n",
      "Remote_IMA_at                     5780 non-null int64\n",
      "dtypes: int64(63)\n",
      "memory usage: 2.8 MB\n"
     ]
    }
   ],
   "source": [
    "dfx = dfx.select_dtypes(exclude=['object', 'datetime64'])\n",
    "dfx = dfx.drop(labels = ['SCORE', 'smoking', 'smoking_recod', 'LDL_calc', 'glucose', 'Unnamed: 0', 'Unnamed: 0.1'], axis=1) \n",
    "dfx = dfx.fillna(-1)\n",
    "dfx = dfx.drop([col for col in dfx.columns if \"date\" in col], axis = 1)\n",
    "dfx = dfx[dfx.columns[dfx.min()>-1]]\n",
    "dfx = dfx[dfx.columns[dfx.max()>0]]\n",
    "dfx.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5780, 63)\n"
     ]
    }
   ],
   "source": [
    "X = dfx.values\n",
    "print(X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(32, activation='sigmoid', input_shape = (X.shape[1],)))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(32, activation='sigmoid'))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(16, activation='sigmoid'))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(4, activation='softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_tr, X_ts, y_tr, y_ts = splt(X, y_cat, test_size=0.25, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 4335 samples, validate on 1445 samples\n",
      "Epoch 1/500\n",
      "4335/4335 [==============================] - 1s 123us/step - loss: 1.4423 - acc: 0.2593 - val_loss: 1.3556 - val_acc: 0.3080\n",
      "Epoch 2/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.3876 - acc: 0.2824 - val_loss: 1.3541 - val_acc: 0.2879\n",
      "Epoch 3/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.3814 - acc: 0.2819 - val_loss: 1.3523 - val_acc: 0.3163\n",
      "Epoch 4/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 1.3737 - acc: 0.2860 - val_loss: 1.3496 - val_acc: 0.4028\n",
      "Epoch 5/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 1.3634 - acc: 0.3040 - val_loss: 1.3400 - val_acc: 0.3723\n",
      "Epoch 6/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.3609 - acc: 0.3246 - val_loss: 1.3323 - val_acc: 0.4353\n",
      "Epoch 7/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.3469 - acc: 0.3398 - val_loss: 1.3074 - val_acc: 0.4457\n",
      "Epoch 8/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.3247 - acc: 0.3822 - val_loss: 1.2810 - val_acc: 0.4671\n",
      "Epoch 9/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 1.2930 - acc: 0.4023 - val_loss: 1.2170 - val_acc: 0.4865\n",
      "Epoch 10/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.2550 - acc: 0.4371 - val_loss: 1.1789 - val_acc: 0.4997\n",
      "Epoch 11/500\n",
      "4335/4335 [==============================] - 0s 33us/step - loss: 1.2407 - acc: 0.4337 - val_loss: 1.1370 - val_acc: 0.5370\n",
      "Epoch 12/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.2055 - acc: 0.4574 - val_loss: 1.1116 - val_acc: 0.5163\n",
      "Epoch 13/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.1628 - acc: 0.4833 - val_loss: 1.0983 - val_acc: 0.5204\n",
      "Epoch 14/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 1.1491 - acc: 0.4939 - val_loss: 1.0342 - val_acc: 0.5682\n",
      "Epoch 15/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.1425 - acc: 0.4937 - val_loss: 1.0532 - val_acc: 0.5412\n",
      "Epoch 16/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.1634 - acc: 0.4812 - val_loss: 1.0348 - val_acc: 0.5405\n",
      "Epoch 17/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.1091 - acc: 0.5045 - val_loss: 1.0230 - val_acc: 0.5474\n",
      "Epoch 18/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 1.0873 - acc: 0.5156 - val_loss: 0.9825 - val_acc: 0.5702\n",
      "Epoch 19/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.0731 - acc: 0.5234 - val_loss: 0.9766 - val_acc: 0.5619\n",
      "Epoch 20/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 1.0682 - acc: 0.5329 - val_loss: 1.0586 - val_acc: 0.5433\n",
      "Epoch 21/500\n",
      "4335/4335 [==============================] - 0s 34us/step - loss: 1.0623 - acc: 0.5326 - val_loss: 1.0135 - val_acc: 0.5564\n",
      "Epoch 22/500\n",
      "4335/4335 [==============================] - 0s 34us/step - loss: 1.0614 - acc: 0.5386 - val_loss: 0.9475 - val_acc: 0.6035\n",
      "Epoch 23/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.0436 - acc: 0.5409 - val_loss: 0.9649 - val_acc: 0.5792\n",
      "Epoch 24/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 1.0314 - acc: 0.5409 - val_loss: 0.9310 - val_acc: 0.5965\n",
      "Epoch 25/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 1.0236 - acc: 0.5444 - val_loss: 0.9433 - val_acc: 0.5896\n",
      "Epoch 26/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 1.0112 - acc: 0.5564 - val_loss: 0.9361 - val_acc: 0.5834\n",
      "Epoch 27/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.0096 - acc: 0.5511 - val_loss: 0.9519 - val_acc: 0.5896\n",
      "Epoch 28/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 1.0145 - acc: 0.5529 - val_loss: 0.9141 - val_acc: 0.5875\n",
      "Epoch 29/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.9868 - acc: 0.5576 - val_loss: 0.9039 - val_acc: 0.5972\n",
      "Epoch 30/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.9915 - acc: 0.5619 - val_loss: 0.8782 - val_acc: 0.6166\n",
      "Epoch 31/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9850 - acc: 0.5647 - val_loss: 0.8732 - val_acc: 0.5993\n",
      "Epoch 32/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9807 - acc: 0.5672 - val_loss: 0.8851 - val_acc: 0.5986\n",
      "Epoch 33/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9679 - acc: 0.5645 - val_loss: 0.8646 - val_acc: 0.6111\n",
      "Epoch 34/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9594 - acc: 0.5790 - val_loss: 0.8630 - val_acc: 0.6021\n",
      "Epoch 35/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9592 - acc: 0.5728 - val_loss: 0.8635 - val_acc: 0.6104\n",
      "Epoch 36/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9513 - acc: 0.5705 - val_loss: 0.8399 - val_acc: 0.6298\n",
      "Epoch 37/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.9428 - acc: 0.5783 - val_loss: 0.8469 - val_acc: 0.6201\n",
      "Epoch 38/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.9595 - acc: 0.5679 - val_loss: 0.8374 - val_acc: 0.6367\n",
      "Epoch 39/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9459 - acc: 0.5839 - val_loss: 0.8147 - val_acc: 0.6235\n",
      "Epoch 40/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9415 - acc: 0.5755 - val_loss: 0.8324 - val_acc: 0.6090\n",
      "Epoch 41/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.9297 - acc: 0.5769 - val_loss: 0.8620 - val_acc: 0.6159\n",
      "Epoch 42/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9239 - acc: 0.5815 - val_loss: 0.8221 - val_acc: 0.6125\n",
      "Epoch 43/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.9068 - acc: 0.5875 - val_loss: 0.8468 - val_acc: 0.6235\n",
      "Epoch 44/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.9213 - acc: 0.5894 - val_loss: 0.7996 - val_acc: 0.6388\n",
      "Epoch 45/500\n",
      "4335/4335 [==============================] - 0s 33us/step - loss: 0.9198 - acc: 0.5970 - val_loss: 0.7888 - val_acc: 0.6519\n",
      "Epoch 46/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.9029 - acc: 0.5938 - val_loss: 0.7986 - val_acc: 0.6415\n",
      "Epoch 47/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.9088 - acc: 0.5988 - val_loss: 0.7999 - val_acc: 0.6512\n",
      "Epoch 48/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.9139 - acc: 0.5880 - val_loss: 0.7936 - val_acc: 0.6526\n",
      "Epoch 49/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9037 - acc: 0.5963 - val_loss: 0.7969 - val_acc: 0.6401\n",
      "Epoch 50/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8799 - acc: 0.6092 - val_loss: 0.7991 - val_acc: 0.6270\n",
      "Epoch 51/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8858 - acc: 0.6090 - val_loss: 0.7825 - val_acc: 0.6457\n",
      "Epoch 52/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8907 - acc: 0.6023 - val_loss: 0.8022 - val_acc: 0.6436\n",
      "Epoch 53/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8881 - acc: 0.6032 - val_loss: 0.7798 - val_acc: 0.6602\n",
      "Epoch 54/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.9239 - acc: 0.5910 - val_loss: 0.7651 - val_acc: 0.6540\n",
      "Epoch 55/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8882 - acc: 0.5977 - val_loss: 0.7734 - val_acc: 0.6547\n",
      "Epoch 56/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8739 - acc: 0.6161 - val_loss: 0.7837 - val_acc: 0.6602\n",
      "Epoch 57/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.8932 - acc: 0.5995 - val_loss: 0.7700 - val_acc: 0.6747\n",
      "Epoch 58/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8949 - acc: 0.6018 - val_loss: 0.7645 - val_acc: 0.6837\n",
      "Epoch 59/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8975 - acc: 0.5931 - val_loss: 0.7703 - val_acc: 0.6616\n",
      "Epoch 60/500\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8894 - acc: 0.6076 - val_loss: 0.7737 - val_acc: 0.6471\n",
      "Epoch 61/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8898 - acc: 0.6053 - val_loss: 0.7755 - val_acc: 0.6360\n",
      "Epoch 62/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8643 - acc: 0.6108 - val_loss: 0.7637 - val_acc: 0.6651\n",
      "Epoch 63/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8860 - acc: 0.6032 - val_loss: 0.7702 - val_acc: 0.6976\n",
      "Epoch 64/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8666 - acc: 0.6189 - val_loss: 0.7605 - val_acc: 0.6740\n",
      "Epoch 65/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8624 - acc: 0.6219 - val_loss: 0.8143 - val_acc: 0.6291\n",
      "Epoch 66/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8588 - acc: 0.6175 - val_loss: 0.7578 - val_acc: 0.6768\n",
      "Epoch 67/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8454 - acc: 0.6175 - val_loss: 0.7557 - val_acc: 0.6740\n",
      "Epoch 68/500\n",
      "4335/4335 [==============================] - 0s 33us/step - loss: 0.8661 - acc: 0.6168 - val_loss: 0.7808 - val_acc: 0.6554\n",
      "Epoch 69/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8677 - acc: 0.6108 - val_loss: 0.7628 - val_acc: 0.6671\n",
      "Epoch 70/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8410 - acc: 0.6182 - val_loss: 0.7494 - val_acc: 0.6747\n",
      "Epoch 71/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8715 - acc: 0.6099 - val_loss: 0.7332 - val_acc: 0.6927\n",
      "Epoch 72/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8444 - acc: 0.6212 - val_loss: 0.7476 - val_acc: 0.6540\n",
      "Epoch 73/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8726 - acc: 0.6088 - val_loss: 0.8264 - val_acc: 0.6090\n",
      "Epoch 74/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8937 - acc: 0.5933 - val_loss: 0.7525 - val_acc: 0.6602\n",
      "Epoch 75/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8449 - acc: 0.6164 - val_loss: 0.7404 - val_acc: 0.6768\n",
      "Epoch 76/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8468 - acc: 0.6201 - val_loss: 0.7506 - val_acc: 0.6844\n",
      "Epoch 77/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8429 - acc: 0.6226 - val_loss: 0.7453 - val_acc: 0.6810\n",
      "Epoch 78/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8454 - acc: 0.6175 - val_loss: 0.7372 - val_acc: 0.6796\n",
      "Epoch 79/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8154 - acc: 0.6360 - val_loss: 0.7375 - val_acc: 0.6913\n",
      "Epoch 80/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8291 - acc: 0.6240 - val_loss: 0.7713 - val_acc: 0.6554\n",
      "Epoch 81/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8446 - acc: 0.6221 - val_loss: 0.7389 - val_acc: 0.6872\n",
      "Epoch 82/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8691 - acc: 0.6035 - val_loss: 0.7737 - val_acc: 0.6879\n",
      "Epoch 83/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8597 - acc: 0.6187 - val_loss: 0.7742 - val_acc: 0.6526\n",
      "Epoch 84/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8611 - acc: 0.6171 - val_loss: 0.7650 - val_acc: 0.6574\n",
      "Epoch 85/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8202 - acc: 0.6245 - val_loss: 0.7414 - val_acc: 0.6581\n",
      "Epoch 86/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8259 - acc: 0.6316 - val_loss: 0.7457 - val_acc: 0.6879\n",
      "Epoch 87/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8275 - acc: 0.6309 - val_loss: 0.7245 - val_acc: 0.6907\n",
      "Epoch 88/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8365 - acc: 0.6284 - val_loss: 0.8513 - val_acc: 0.6076\n",
      "Epoch 89/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8800 - acc: 0.5933 - val_loss: 0.7405 - val_acc: 0.6747\n",
      "Epoch 90/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8660 - acc: 0.6187 - val_loss: 0.7310 - val_acc: 0.6796\n",
      "Epoch 91/500\n",
      "4335/4335 [==============================] - 0s 35us/step - loss: 0.8456 - acc: 0.6143 - val_loss: 0.7742 - val_acc: 0.6747\n",
      "Epoch 92/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.8214 - acc: 0.6371 - val_loss: 0.7342 - val_acc: 0.6768\n",
      "Epoch 93/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8362 - acc: 0.6238 - val_loss: 0.8578 - val_acc: 0.6021\n",
      "Epoch 94/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8231 - acc: 0.6203 - val_loss: 0.7424 - val_acc: 0.6920\n",
      "Epoch 95/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8136 - acc: 0.6187 - val_loss: 0.7665 - val_acc: 0.6754\n",
      "Epoch 96/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8127 - acc: 0.6337 - val_loss: 0.7665 - val_acc: 0.6706\n",
      "Epoch 97/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8454 - acc: 0.6168 - val_loss: 0.8032 - val_acc: 0.6581\n",
      "Epoch 98/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8434 - acc: 0.6164 - val_loss: 0.7329 - val_acc: 0.6817\n",
      "Epoch 99/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8142 - acc: 0.6286 - val_loss: 0.7235 - val_acc: 0.6886\n",
      "Epoch 100/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8232 - acc: 0.6208 - val_loss: 0.7655 - val_acc: 0.6768\n",
      "Epoch 101/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8166 - acc: 0.6302 - val_loss: 0.7409 - val_acc: 0.6824\n",
      "Epoch 102/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8348 - acc: 0.6231 - val_loss: 0.7396 - val_acc: 0.6734\n",
      "Epoch 103/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8267 - acc: 0.6293 - val_loss: 0.7231 - val_acc: 0.6941\n",
      "Epoch 104/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8211 - acc: 0.6279 - val_loss: 0.7338 - val_acc: 0.6872\n",
      "Epoch 105/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7973 - acc: 0.6399 - val_loss: 0.7397 - val_acc: 0.6747\n",
      "Epoch 106/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8063 - acc: 0.6489 - val_loss: 0.7467 - val_acc: 0.6491\n",
      "Epoch 107/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8232 - acc: 0.6208 - val_loss: 0.7323 - val_acc: 0.6872\n",
      "Epoch 108/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8019 - acc: 0.6408 - val_loss: 0.7165 - val_acc: 0.6844\n",
      "Epoch 109/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8049 - acc: 0.6330 - val_loss: 0.7026 - val_acc: 0.6879\n",
      "Epoch 110/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8626 - acc: 0.6187 - val_loss: 0.7347 - val_acc: 0.6879\n",
      "Epoch 111/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8279 - acc: 0.6394 - val_loss: 0.7215 - val_acc: 0.6844\n",
      "Epoch 112/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8324 - acc: 0.6245 - val_loss: 0.7118 - val_acc: 0.6775\n",
      "Epoch 113/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7994 - acc: 0.6411 - val_loss: 0.7112 - val_acc: 0.6900\n",
      "Epoch 114/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8174 - acc: 0.6337 - val_loss: 0.7130 - val_acc: 0.6907\n",
      "Epoch 115/500\n",
      "4335/4335 [==============================] - 0s 33us/step - loss: 0.8559 - acc: 0.6201 - val_loss: 0.7319 - val_acc: 0.6789\n",
      "Epoch 116/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8470 - acc: 0.6157 - val_loss: 0.7062 - val_acc: 0.6879\n",
      "Epoch 117/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8061 - acc: 0.6378 - val_loss: 0.7255 - val_acc: 0.6851\n",
      "Epoch 118/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8011 - acc: 0.6291 - val_loss: 0.7103 - val_acc: 0.6893\n",
      "Epoch 119/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8081 - acc: 0.6420 - val_loss: 0.7461 - val_acc: 0.6754\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 120/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8150 - acc: 0.6247 - val_loss: 0.7052 - val_acc: 0.6803\n",
      "Epoch 121/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8074 - acc: 0.6378 - val_loss: 0.7356 - val_acc: 0.6851\n",
      "Epoch 122/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8107 - acc: 0.6330 - val_loss: 0.7495 - val_acc: 0.6796\n",
      "Epoch 123/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8095 - acc: 0.6464 - val_loss: 0.7069 - val_acc: 0.6913\n",
      "Epoch 124/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8282 - acc: 0.6311 - val_loss: 0.7129 - val_acc: 0.6858\n",
      "Epoch 125/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.8141 - acc: 0.6298 - val_loss: 0.7104 - val_acc: 0.6941\n",
      "Epoch 126/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8093 - acc: 0.6424 - val_loss: 0.6938 - val_acc: 0.6893\n",
      "Epoch 127/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8457 - acc: 0.6311 - val_loss: 0.7070 - val_acc: 0.6886\n",
      "Epoch 128/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8203 - acc: 0.6323 - val_loss: 0.7070 - val_acc: 0.6893\n",
      "Epoch 129/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8003 - acc: 0.6385 - val_loss: 0.6913 - val_acc: 0.6927\n",
      "Epoch 130/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7988 - acc: 0.6429 - val_loss: 0.7528 - val_acc: 0.6851\n",
      "Epoch 131/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7942 - acc: 0.6411 - val_loss: 0.7127 - val_acc: 0.6920\n",
      "Epoch 132/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8075 - acc: 0.6390 - val_loss: 0.7059 - val_acc: 0.6844\n",
      "Epoch 133/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8240 - acc: 0.6369 - val_loss: 0.7136 - val_acc: 0.6886\n",
      "Epoch 134/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7789 - acc: 0.6496 - val_loss: 0.7066 - val_acc: 0.6851\n",
      "Epoch 135/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.7932 - acc: 0.6480 - val_loss: 0.7199 - val_acc: 0.6893\n",
      "Epoch 136/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8117 - acc: 0.6415 - val_loss: 0.6920 - val_acc: 0.6851\n",
      "Epoch 137/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8088 - acc: 0.6298 - val_loss: 0.7048 - val_acc: 0.6810\n",
      "Epoch 138/500\n",
      "4335/4335 [==============================] - 0s 34us/step - loss: 0.7831 - acc: 0.6514 - val_loss: 0.8103 - val_acc: 0.6623\n",
      "Epoch 139/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.8110 - acc: 0.6537 - val_loss: 0.7154 - val_acc: 0.6817\n",
      "Epoch 140/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8034 - acc: 0.6401 - val_loss: 0.7279 - val_acc: 0.6920\n",
      "Epoch 141/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7850 - acc: 0.6507 - val_loss: 0.7072 - val_acc: 0.6886\n",
      "Epoch 142/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7900 - acc: 0.6452 - val_loss: 0.7096 - val_acc: 0.6941\n",
      "Epoch 143/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7861 - acc: 0.6533 - val_loss: 0.7313 - val_acc: 0.6671\n",
      "Epoch 144/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7961 - acc: 0.6351 - val_loss: 0.6965 - val_acc: 0.6900\n",
      "Epoch 145/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7951 - acc: 0.6464 - val_loss: 0.7030 - val_acc: 0.6768\n",
      "Epoch 146/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7775 - acc: 0.6461 - val_loss: 0.7171 - val_acc: 0.6796\n",
      "Epoch 147/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8218 - acc: 0.6304 - val_loss: 0.7221 - val_acc: 0.6837\n",
      "Epoch 148/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.8032 - acc: 0.6404 - val_loss: 0.7011 - val_acc: 0.6837\n",
      "Epoch 149/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8071 - acc: 0.6464 - val_loss: 0.6868 - val_acc: 0.6886\n",
      "Epoch 150/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7862 - acc: 0.6521 - val_loss: 0.6852 - val_acc: 0.6934\n",
      "Epoch 151/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7979 - acc: 0.6351 - val_loss: 0.7223 - val_acc: 0.6734\n",
      "Epoch 152/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8094 - acc: 0.6291 - val_loss: 0.7011 - val_acc: 0.6865\n",
      "Epoch 153/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7869 - acc: 0.6468 - val_loss: 0.7018 - val_acc: 0.6913\n",
      "Epoch 154/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8232 - acc: 0.6399 - val_loss: 0.7133 - val_acc: 0.6920\n",
      "Epoch 155/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.8086 - acc: 0.6383 - val_loss: 0.7206 - val_acc: 0.6844\n",
      "Epoch 156/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7875 - acc: 0.6489 - val_loss: 0.6972 - val_acc: 0.6775\n",
      "Epoch 157/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.7771 - acc: 0.6558 - val_loss: 0.7106 - val_acc: 0.6983\n",
      "Epoch 158/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7865 - acc: 0.6401 - val_loss: 0.7450 - val_acc: 0.6900\n",
      "Epoch 159/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8039 - acc: 0.6390 - val_loss: 0.7025 - val_acc: 0.6893\n",
      "Epoch 160/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.8072 - acc: 0.6429 - val_loss: 0.7015 - val_acc: 0.6796\n",
      "Epoch 161/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.7926 - acc: 0.6450 - val_loss: 0.7929 - val_acc: 0.6512\n",
      "Epoch 162/500\n",
      "4335/4335 [==============================] - 0s 32us/step - loss: 0.7795 - acc: 0.6521 - val_loss: 0.7088 - val_acc: 0.6886\n",
      "Epoch 163/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.8041 - acc: 0.6358 - val_loss: 0.7014 - val_acc: 0.6768\n",
      "Epoch 164/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7980 - acc: 0.6390 - val_loss: 0.7391 - val_acc: 0.6727\n",
      "Epoch 165/500\n",
      "4335/4335 [==============================] - 0s 31us/step - loss: 0.7928 - acc: 0.6459 - val_loss: 0.6979 - val_acc: 0.6886\n",
      "Epoch 166/500\n",
      "4335/4335 [==============================] - 0s 29us/step - loss: 0.7876 - acc: 0.6452 - val_loss: 0.6986 - val_acc: 0.6893\n",
      "Epoch 167/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7950 - acc: 0.6448 - val_loss: 0.7528 - val_acc: 0.6796\n",
      "Epoch 168/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8097 - acc: 0.6408 - val_loss: 0.7142 - val_acc: 0.6934\n",
      "Epoch 169/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.8058 - acc: 0.6434 - val_loss: 0.7086 - val_acc: 0.6872\n",
      "Epoch 170/500\n",
      "4335/4335 [==============================] - 0s 30us/step - loss: 0.7970 - acc: 0.6420 - val_loss: 0.7407 - val_acc: 0.6685\n"
     ]
    }
   ],
   "source": [
    "es = EarlyStopping(patience = 20, restore_best_weights = True)\n",
    "history = model.fit(x=X_tr, y=y_tr, validation_data=(X_ts, y_ts), epochs=500, callbacks = [es], class_weight = {0: 1, 1: 1, 2: 1, 3: 1})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_history(history, filename, filename2):\n",
    "    loss_list = [s for s in history.history.keys() if 'loss' in s and 'val' not in s]\n",
    "    val_loss_list = [s for s in history.history.keys() if 'loss' in s and 'val' in s]\n",
    "    acc_list = [s for s in history.history.keys() if 'acc' in s and 'val' not in s]\n",
    "    val_acc_list = [s for s in history.history.keys() if 'acc' in s and 'val' in s]\n",
    "    \n",
    "    if len(loss_list) == 0:\n",
    "        print('Loss is missing in history')\n",
    "        return \n",
    "    \n",
    "    ## As loss always exists\n",
    "    epochs = range(1,len(history.history[loss_list[0]]) + 1)\n",
    "    \n",
    "    ## Loss\n",
    "    plt.figure(1)\n",
    "    for l in loss_list:\n",
    "        plt.plot(epochs, history.history[l], 'b', label='Training loss (' + str(str(format(history.history[l][-1],'.5f'))+')'))\n",
    "    for l in val_loss_list:\n",
    "        plt.plot(epochs, history.history[l], 'g', label='Validation loss (' + str(str(format(history.history[l][-1],'.5f'))+')'))\n",
    "    \n",
    "    plt.title('Loss')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.ylabel('Loss')\n",
    "    plt.legend(loc=\"lower right\")\n",
    "    plt.savefig(filename)\n",
    "    plt.ylim(0, 1.5)\n",
    "    \n",
    "    ## Accuracy\n",
    "    plt.figure(2)\n",
    "    for l in acc_list:\n",
    "        plt.plot(epochs, history.history[l], 'b', label='Training accuracy (' + str(format(history.history[l][-1],'.5f'))+')')\n",
    "    for l in val_acc_list:    \n",
    "        plt.plot(epochs, history.history[l], 'g', label='Validation accuracy (' + str(format(history.history[l][-1],'.5f'))+')')\n",
    "\n",
    "    plt.title('Accuracy')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.legend(loc=\"lower right\")\n",
    "    plt.ylim(0,1)\n",
    "    plt.savefig(filename2)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_history(history, \"Graphs/BNE_2_LOSS.png\", \"Graphs/BNE_2_ACC.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_ts_pred = model.predict(X_ts)\n",
    "cmatrix = np.zeros((4,4))\n",
    "for i,y_our in enumerate(y_ts_pred):\n",
    "    cmatrix[np.argmax(y_our)][np.argmax(y_ts[i])] += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_cm = pd.DataFrame(cmatrix, index = [\"No\", \"Low\", \"Med\", \"High\"], columns = [\"No\", \"Low\", \"Med\", \"High\"])\n",
    "plt.figure(figsize = (10,7))\n",
    "snheatmap = sn.heatmap(df_cm, annot=True,  fmt='g', cmap=\"Blues\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "snheatmap.get_figure().savefig(\"BNE_2_CONFUSION_MATRIX.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}