harml_nn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation\n",
    "from keras.utils import to_categorical\n",
    "from keras.optimizers import Adam\n",
    "from sklearn import metrics\n",
    "from sklearn.preprocessing import normalize\n",
    "import matplotlib.pyplot as plt\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "INPUT_DATA = [\n",
    "    \"body_acc_x_\",\n",
    "    \"body_acc_y_\",\n",
    "    \"body_acc_z_\",\n",
    "    \"body_gyro_x_\",\n",
    "    \"body_gyro_y_\",\n",
    "    \"body_gyro_z_\",\n",
    "    \"total_acc_x_\",\n",
    "    \"total_acc_y_\",\n",
    "    \"total_acc_z_\"\n",
    "]\n",
    "\n",
    "LABELS = [\n",
    "    \"WALKING\", \n",
    "    \"WALKING_UPSTAIRS\", \n",
    "    \"WALKING_DOWNSTAIRS\", \n",
    "    \"SITTING\", \n",
    "    \"STANDING\", \n",
    "    \"LAYING\"\n",
    "] \n",
    "\n",
    "TRAINING_SIZE = 7000\n",
    "TESTING_SIZE = 2500"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def feature_extraction(x, y, z):\n",
    "    features = [np.mean(x), np.mean(y), np.mean(z), np.std(x), np.std(y), np.std(z)]\n",
    "    #Median Absolute Deviation\n",
    "    features.extend((np.mean(abs(x - features[0])), np.mean(abs(y - features[1])), np.mean(abs(z - features[2]))))\n",
    "    #Jerk Signals\n",
    "    features.extend((np.mean(np.diff(x)), np.mean(np.diff(y)), np.mean(np.diff(z)), np.std(np.diff(x)), np.std(np.diff(y)), np.std(np.diff(z))))\n",
    "    features.extend((np.mean(abs(np.diff(x) - features[9])), np.mean(abs(np.diff(y) - features[10])), np.mean(abs(np.diff(y) - features[11]))))\n",
    "    #features.extend((max(x), max(y), max(z), min(x), min(y), min(z)))\n",
    "    return features\n",
    "\n",
    "def feature_selection(X):\n",
    "    data = []\n",
    "    for i in range(X.shape[0]):\n",
    "        features = []\n",
    "        for j in range(0, X.shape[2], 3):\n",
    "            x = [X[i][u][j] for u in range(X.shape[1])]\n",
    "            y = [X[i][u][j+1] for u in range(X.shape[1])]\n",
    "            z = [X[i][u][j+2] for u in range(X.shape[1])]\n",
    "            features.append(feature_extraction(x, y, z))\n",
    "        data.append(features)\n",
    "        #data.append(normalize(features))\n",
    "    return np.array(data)\n",
    "\n",
    "def load_X(X_signals_paths):\n",
    "    X_signals = []\n",
    "    \n",
    "    for signal_type_path in X_signals_paths:\n",
    "        file = open(signal_type_path, 'r')\n",
    "        # Read dataset from disk, dealing with text files' syntax\n",
    "        X_signals.append(\n",
    "            [np.array(serie, dtype=np.float32) for serie in \n",
    "             [row.replace('  ', ' ').strip().split(' ') for row in file]\n",
    "            ]\n",
    "        )\n",
    "        file.close()\n",
    "    \n",
    "    return np.transpose(np.array(X_signals), (1, 2, 0))\n",
    "\n",
    "\n",
    "def load_y(y_path):\n",
    "    file = open(y_path, 'r')\n",
    "    # Read dataset from disk, dealing with text file's syntax\n",
    "    y_ = np.array(\n",
    "        [elem for elem in [\n",
    "            row.replace('  ', ' ').strip().split(' ') for row in file\n",
    "        ]], \n",
    "        dtype=np.int32\n",
    "    )\n",
    "    file.close()\n",
    "    return y_ - 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7352, 128, 9)\n",
      "(7352,)\n",
      "(2947, 128, 9)\n",
      "(2947,)\n",
      "(7000, 128, 9)\n",
      "(7000,)\n",
      "(2500, 128, 9)\n",
      "(2500,)\n",
      "(7000, 3, 18)\n",
      "(2500, 3, 18)\n",
      "(7000, 6)\n",
      "[[  2.26869062e-03  -1.65537116e-03  -8.12555104e-03   2.94102891e-03\n",
      "    4.60070046e-03   1.85043626e-02   2.51075346e-03   3.61979101e-03\n",
      "    1.32809971e-02   1.09814628e-05  -6.06032881e-05  -4.55357687e-04\n",
      "    2.00701482e-03   2.24945950e-03   2.90326332e-03   1.43474573e-03\n",
      "    1.77149521e-03   1.79056870e-03]\n",
      " [  2.22503506e-02   1.94027703e-02   9.46815498e-03   1.44340890e-02\n",
      "    1.98283661e-02   5.94323827e-03   1.23813283e-02   1.64496675e-02\n",
      "    4.78596706e-03  -1.08142685e-05  -5.20089234e-04  -1.91375191e-04\n",
      "    4.00003418e-03   5.96561469e-03   3.71990330e-03   3.02015617e-03\n",
      "    3.50187998e-03   3.47918598e-03]\n",
      " [  1.01930463e+00  -1.24261871e-01   9.97317880e-02   1.79055484e-03\n",
      "    3.26999812e-03   3.96959949e-03   1.33998133e-03   2.65987287e-03\n",
      "    3.00057116e-03   6.57791243e-05   1.48874333e-05  -6.25723842e-05\n",
      "    2.00633937e-03   2.23925756e-03   2.66895629e-03   1.43639895e-03\n",
      "    1.76221225e-03   1.76160212e-03]]\n",
      "(7000, 54)\n",
      "(2500, 54)\n",
      "[  2.26869062e-03  -1.65537116e-03  -8.12555104e-03   2.94102891e-03\n",
      "   4.60070046e-03   1.85043626e-02   2.51075346e-03   3.61979101e-03\n",
      "   1.32809971e-02   1.09814628e-05  -6.06032881e-05  -4.55357687e-04\n",
      "   2.00701482e-03   2.24945950e-03   2.90326332e-03   1.43474573e-03\n",
      "   1.77149521e-03   1.79056870e-03   2.22503506e-02   1.94027703e-02\n",
      "   9.46815498e-03   1.44340890e-02   1.98283661e-02   5.94323827e-03\n",
      "   1.23813283e-02   1.64496675e-02   4.78596706e-03  -1.08142685e-05\n",
      "  -5.20089234e-04  -1.91375191e-04   4.00003418e-03   5.96561469e-03\n",
      "   3.71990330e-03   3.02015617e-03   3.50187998e-03   3.47918598e-03\n",
      "   1.01930463e+00  -1.24261871e-01   9.97317880e-02   1.79055484e-03\n",
      "   3.26999812e-03   3.96959949e-03   1.33998133e-03   2.65987287e-03\n",
      "   3.00057116e-03   6.57791243e-05   1.48874333e-05  -6.25723842e-05\n",
      "   2.00633937e-03   2.23925756e-03   2.66895629e-03   1.43639895e-03\n",
      "   1.76221225e-03   1.76160212e-03]\n"
     ]
    }
   ],
   "source": [
    "\n",
    "X_train_input_paths = [\"data/train/\" + signal + \"train.txt\" for signal in INPUT_DATA]\n",
    "X_test_input_paths = [\"data/test/\" + signal + \"test.txt\" for signal in INPUT_DATA]\n",
    "X_train = load_X(X_train_input_paths)\n",
    "X_test = load_X(X_test_input_paths)\n",
    "\n",
    "y_train_path = \"data/y_train.txt\"\n",
    "y_test_path = \"data/y_test.txt\"\n",
    "Y_train = load_y(y_train_path).ravel()\n",
    "Y_test = load_y(y_test_path).ravel()\n",
    "\n",
    "print(X_train.shape)\n",
    "print(Y_train.shape)\n",
    "print(X_test.shape)\n",
    "print(Y_test.shape)\n",
    "\n",
    "#print(X_train[0])\n",
    "#print(Y_train[0])\n",
    "\n",
    "X = X_train[0:TRAINING_SIZE]\n",
    "Y = Y_train[0:TRAINING_SIZE]\n",
    "Xt = X_test[0:TESTING_SIZE]\n",
    "Yt = Y_test[0:TESTING_SIZE]\n",
    "\n",
    "print(X.shape)\n",
    "print(Y.shape)\n",
    "print(Xt.shape)\n",
    "print(Yt.shape)\n",
    "\n",
    "\n",
    "X = feature_selection(X)\n",
    "Xt = feature_selection(Xt)\n",
    "Yhot = to_categorical(Y)\n",
    "\n",
    "print(X.shape)\n",
    "print(Xt.shape)\n",
    "print(Yhot.shape)\n",
    "print(X[0])\n",
    "\n",
    "np.savetxt(\"data/test_data_format_raw.csv\", X[0], delimiter=\",\")\n",
    "\n",
    "X = X.reshape(X.shape[0], X.shape[1]*X.shape[2])\n",
    "Xt = Xt.reshape(Xt.shape[0], Xt.shape[1]*Xt.shape[2])\n",
    "\n",
    "#X = normalize(X)\n",
    "#Xt = normalize(Xt)\n",
    "\n",
    "print(X.shape)\n",
    "print(Xt.shape)\n",
    "print(X[0])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/arshanrahman/anaconda/lib/python3.6/site-packages/ipykernel_launcher.py:4: UserWarning: Update your `Dense` call to the Keras 2 API: `Dense(64, input_dim=54, activation=\"relu\", kernel_initializer=\"normal\")`\n",
      "  after removing the cwd from sys.path.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 5600 samples, validate on 1400 samples\n",
      "Epoch 1/100\n",
      "5600/5600 [==============================] - 0s - loss: 1.3886 - acc: 0.5500 - val_loss: 0.8624 - val_acc: 0.7486\n",
      "Epoch 2/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.6629 - acc: 0.7268 - val_loss: 0.5366 - val_acc: 0.8407\n",
      "Epoch 3/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.4838 - acc: 0.7943 - val_loss: 0.4566 - val_acc: 0.8679\n",
      "Epoch 4/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.4043 - acc: 0.8370 - val_loss: 0.3945 - val_acc: 0.8907\n",
      "Epoch 5/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.3461 - acc: 0.8596 - val_loss: 0.3495 - val_acc: 0.8993\n",
      "Epoch 6/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2980 - acc: 0.8812 - val_loss: 0.3273 - val_acc: 0.9014\n",
      "Epoch 7/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2722 - acc: 0.8861 - val_loss: 0.2992 - val_acc: 0.9071\n",
      "Epoch 8/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2656 - acc: 0.8929 - val_loss: 0.2875 - val_acc: 0.9171\n",
      "Epoch 9/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2513 - acc: 0.9018 - val_loss: 0.2916 - val_acc: 0.9136\n",
      "Epoch 10/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2400 - acc: 0.9059 - val_loss: 0.3042 - val_acc: 0.9064\n",
      "Epoch 11/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2333 - acc: 0.9100 - val_loss: 0.2819 - val_acc: 0.9193\n",
      "Epoch 12/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2250 - acc: 0.9123 - val_loss: 0.3037 - val_acc: 0.9114\n",
      "Epoch 13/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2193 - acc: 0.9093 - val_loss: 0.2897 - val_acc: 0.9143\n",
      "Epoch 14/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2121 - acc: 0.9132 - val_loss: 0.2847 - val_acc: 0.9136\n",
      "Epoch 15/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2156 - acc: 0.9123 - val_loss: 0.2867 - val_acc: 0.9121\n",
      "Epoch 16/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.2043 - acc: 0.9177 - val_loss: 0.2739 - val_acc: 0.9186\n",
      "Epoch 17/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1981 - acc: 0.9218 - val_loss: 0.2833 - val_acc: 0.9143\n",
      "Epoch 18/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1933 - acc: 0.9230 - val_loss: 0.2796 - val_acc: 0.9157\n",
      "Epoch 19/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1807 - acc: 0.9273 - val_loss: 0.2859 - val_acc: 0.9164\n",
      "Epoch 20/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1820 - acc: 0.9246 - val_loss: 0.2929 - val_acc: 0.9129\n",
      "Epoch 21/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1902 - acc: 0.9266 - val_loss: 0.2973 - val_acc: 0.9093\n",
      "Epoch 22/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1734 - acc: 0.9323 - val_loss: 0.3185 - val_acc: 0.9043\n",
      "Epoch 23/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1725 - acc: 0.9307 - val_loss: 0.2767 - val_acc: 0.9200\n",
      "Epoch 24/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1685 - acc: 0.9311 - val_loss: 0.2864 - val_acc: 0.9186\n",
      "Epoch 25/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1709 - acc: 0.9321 - val_loss: 0.3009 - val_acc: 0.9150\n",
      "Epoch 26/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1686 - acc: 0.9295 - val_loss: 0.2917 - val_acc: 0.9207\n",
      "Epoch 27/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1636 - acc: 0.9346 - val_loss: 0.2949 - val_acc: 0.9193\n",
      "Epoch 28/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1553 - acc: 0.9368 - val_loss: 0.3063 - val_acc: 0.9143\n",
      "Epoch 29/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1591 - acc: 0.9363 - val_loss: 0.2875 - val_acc: 0.9186\n",
      "Epoch 30/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1508 - acc: 0.9386 - val_loss: 0.2907 - val_acc: 0.9236\n",
      "Epoch 31/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1508 - acc: 0.9407 - val_loss: 0.3034 - val_acc: 0.9150\n",
      "Epoch 32/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1451 - acc: 0.9421 - val_loss: 0.2882 - val_acc: 0.9236\n",
      "Epoch 33/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1475 - acc: 0.9421 - val_loss: 0.2902 - val_acc: 0.9179\n",
      "Epoch 34/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1443 - acc: 0.9413 - val_loss: 0.3089 - val_acc: 0.9143\n",
      "Epoch 35/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1374 - acc: 0.9446 - val_loss: 0.2902 - val_acc: 0.9257\n",
      "Epoch 36/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1453 - acc: 0.9418 - val_loss: 0.3033 - val_acc: 0.9186\n",
      "Epoch 37/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1390 - acc: 0.9416 - val_loss: 0.3024 - val_acc: 0.9250\n",
      "Epoch 38/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1371 - acc: 0.9423 - val_loss: 0.3184 - val_acc: 0.9143\n",
      "Epoch 39/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1384 - acc: 0.9407 - val_loss: 0.3121 - val_acc: 0.9186\n",
      "Epoch 40/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1358 - acc: 0.9446 - val_loss: 0.3272 - val_acc: 0.9107\n",
      "Epoch 41/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1306 - acc: 0.9446 - val_loss: 0.3033 - val_acc: 0.9229\n",
      "Epoch 42/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1315 - acc: 0.9464 - val_loss: 0.3095 - val_acc: 0.9179\n",
      "Epoch 43/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1286 - acc: 0.9464 - val_loss: 0.3096 - val_acc: 0.9293\n",
      "Epoch 44/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1238 - acc: 0.9484 - val_loss: 0.3028 - val_acc: 0.9264\n",
      "Epoch 45/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1304 - acc: 0.9466 - val_loss: 0.3448 - val_acc: 0.9064\n",
      "Epoch 46/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1340 - acc: 0.9448 - val_loss: 0.3127 - val_acc: 0.9171\n",
      "Epoch 47/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1249 - acc: 0.9468 - val_loss: 0.3242 - val_acc: 0.9207\n",
      "Epoch 48/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1246 - acc: 0.9457 - val_loss: 0.3310 - val_acc: 0.9171\n",
      "Epoch 49/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1184 - acc: 0.9500 - val_loss: 0.3340 - val_acc: 0.9150\n",
      "Epoch 50/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1181 - acc: 0.9507 - val_loss: 0.3200 - val_acc: 0.9171\n",
      "Epoch 51/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1216 - acc: 0.9507 - val_loss: 0.3057 - val_acc: 0.9200\n",
      "Epoch 52/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1204 - acc: 0.9543 - val_loss: 0.3111 - val_acc: 0.9179\n",
      "Epoch 53/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1178 - acc: 0.9504 - val_loss: 0.2995 - val_acc: 0.9207\n",
      "Epoch 54/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1187 - acc: 0.9491 - val_loss: 0.3129 - val_acc: 0.9164\n",
      "Epoch 55/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1171 - acc: 0.9505 - val_loss: 0.3224 - val_acc: 0.9171\n",
      "Epoch 56/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1176 - acc: 0.9495 - val_loss: 0.3234 - val_acc: 0.9157\n",
      "Epoch 57/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1109 - acc: 0.9518 - val_loss: 0.3236 - val_acc: 0.9129\n",
      "Epoch 58/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1218 - acc: 0.9487 - val_loss: 0.3339 - val_acc: 0.9143\n",
      "Epoch 59/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1104 - acc: 0.9520 - val_loss: 0.3467 - val_acc: 0.9150\n",
      "Epoch 60/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1147 - acc: 0.9496 - val_loss: 0.3230 - val_acc: 0.9186\n",
      "Epoch 61/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1141 - acc: 0.9525 - val_loss: 0.3309 - val_acc: 0.9143\n",
      "Epoch 62/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1052 - acc: 0.9559 - val_loss: 0.3321 - val_acc: 0.9157\n",
      "Epoch 63/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1108 - acc: 0.9516 - val_loss: 0.3503 - val_acc: 0.9121\n",
      "Epoch 64/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1055 - acc: 0.9554 - val_loss: 0.3344 - val_acc: 0.9171\n",
      "Epoch 65/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5600/5600 [==============================] - 0s - loss: 0.1088 - acc: 0.9530 - val_loss: 0.3279 - val_acc: 0.9164\n",
      "Epoch 66/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1088 - acc: 0.9529 - val_loss: 0.3186 - val_acc: 0.9279\n",
      "Epoch 67/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1118 - acc: 0.9545 - val_loss: 0.3411 - val_acc: 0.9214\n",
      "Epoch 68/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1014 - acc: 0.9568 - val_loss: 0.3445 - val_acc: 0.9221\n",
      "Epoch 69/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1046 - acc: 0.9534 - val_loss: 0.3382 - val_acc: 0.9200\n",
      "Epoch 70/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1033 - acc: 0.9546 - val_loss: 0.3503 - val_acc: 0.9164\n",
      "Epoch 71/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1040 - acc: 0.9557 - val_loss: 0.3512 - val_acc: 0.9221\n",
      "Epoch 72/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1004 - acc: 0.9584 - val_loss: 0.3562 - val_acc: 0.9243\n",
      "Epoch 73/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1022 - acc: 0.9546 - val_loss: 0.3705 - val_acc: 0.9179\n",
      "Epoch 74/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1041 - acc: 0.9577 - val_loss: 0.3678 - val_acc: 0.9207\n",
      "Epoch 75/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0983 - acc: 0.9552 - val_loss: 0.3632 - val_acc: 0.9179\n",
      "Epoch 76/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0971 - acc: 0.9586 - val_loss: 0.3592 - val_acc: 0.9193\n",
      "Epoch 77/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1021 - acc: 0.9573 - val_loss: 0.3642 - val_acc: 0.9214\n",
      "Epoch 78/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0958 - acc: 0.9591 - val_loss: 0.3884 - val_acc: 0.9186\n",
      "Epoch 79/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.1013 - acc: 0.9557 - val_loss: 0.3621 - val_acc: 0.9193\n",
      "Epoch 80/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0954 - acc: 0.9596 - val_loss: 0.3839 - val_acc: 0.9150\n",
      "Epoch 81/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0902 - acc: 0.9607 - val_loss: 0.4024 - val_acc: 0.9150\n",
      "Epoch 82/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0939 - acc: 0.9595 - val_loss: 0.3680 - val_acc: 0.9236\n",
      "Epoch 83/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0962 - acc: 0.9602 - val_loss: 0.3701 - val_acc: 0.9200\n",
      "Epoch 84/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0932 - acc: 0.9602 - val_loss: 0.3680 - val_acc: 0.9200\n",
      "Epoch 85/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0945 - acc: 0.9609 - val_loss: 0.3661 - val_acc: 0.9193\n",
      "Epoch 86/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0917 - acc: 0.9595 - val_loss: 0.3862 - val_acc: 0.9164\n",
      "Epoch 87/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0921 - acc: 0.9632 - val_loss: 0.4036 - val_acc: 0.9143\n",
      "Epoch 88/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0898 - acc: 0.9623 - val_loss: 0.4138 - val_acc: 0.9100\n",
      "Epoch 89/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0886 - acc: 0.9621 - val_loss: 0.4046 - val_acc: 0.9107\n",
      "Epoch 90/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0984 - acc: 0.9577 - val_loss: 0.4033 - val_acc: 0.9057\n",
      "Epoch 91/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0892 - acc: 0.9621 - val_loss: 0.3990 - val_acc: 0.9221\n",
      "Epoch 92/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0858 - acc: 0.9636 - val_loss: 0.4023 - val_acc: 0.9164\n",
      "Epoch 93/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0868 - acc: 0.9641 - val_loss: 0.4032 - val_acc: 0.9143\n",
      "Epoch 94/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0887 - acc: 0.9630 - val_loss: 0.3669 - val_acc: 0.9221\n",
      "Epoch 95/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0888 - acc: 0.9632 - val_loss: 0.4151 - val_acc: 0.9193\n",
      "Epoch 96/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0870 - acc: 0.9613 - val_loss: 0.4256 - val_acc: 0.9236\n",
      "Epoch 97/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0853 - acc: 0.9677 - val_loss: 0.4013 - val_acc: 0.9171\n",
      "Epoch 98/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0839 - acc: 0.9641 - val_loss: 0.4259 - val_acc: 0.9136\n",
      "Epoch 99/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0843 - acc: 0.9648 - val_loss: 0.3941 - val_acc: 0.9229\n",
      "Epoch 100/100\n",
      "5600/5600 [==============================] - 0s - loss: 0.0844 - acc: 0.9630 - val_loss: 0.4022 - val_acc: 0.9236\n",
      "5760/7000 [=======================>......] - ETA: 0s\n",
      "loss: 0.13%\n",
      "\n",
      "acc: 96.41%\n"
     ]
    }
   ],
   "source": [
    "#Neural Network\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(64, input_dim=X.shape[1], init='normal', activation='relu'))\n",
    "model.add(Dropout(0.2))\n",
    "#model.add(BatchNormalization())\n",
    "model.add(Dense(32, activation='relu'))\n",
    "#model.add(Dropout(0.2))\n",
    "#model.add(BatchNormalization())\n",
    "model.add(Dense(16, activation='relu'))\n",
    "#model.add(Dropout(0.2))\n",
    "#model.add(BatchNormalization())\n",
    "model.add(Dense(6, activation='softmax'))\n",
    "\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.fit(X, Yhot, epochs=100, batch_size=64, validation_split=0.2, shuffle=True)\n",
    "result = model.evaluate(X, Yhot)\n",
    "print(\"\\n%s: %.2f%%\" % (model.metrics_names[0], result[0]))\n",
    "print(\"\\n%s: %.2f%%\" % (model.metrics_names[1], result[1]*100))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[4 4 4 ..., 0 1 2]\n",
      "Accuracy Rate:\n",
      "0.8816\n",
      "[[365   3  52   0   0   0]\n",
      " [ 21 342  24   0   0   0]\n",
      " [ 18  13 311   0   0   0]\n",
      " [  0   5   0 314 100   4]\n",
      " [  1   0   0  26 436   0]\n",
      " [  0  29   0   0   0 436]]\n",
      "Precision: 88.76154217428632%\n",
      "Recall: 88.16000000000001%\n",
      "f1_score: 88.14341295099423%\n",
      "\n",
      "Confusion Matrix:\n",
      "[[365   3  52   0   0   0]\n",
      " [ 21 342  24   0   0   0]\n",
      " [ 18  13 311   0   0   0]\n",
      " [  0   5   0 314 100   4]\n",
      " [  1   0   0  26 436   0]\n",
      " [  0  29   0   0   0 436]]\n",
      "\n",
      "Confusion matrix (normalised to % of total test data):\n",
      "[[ 14.59999943   0.12         2.07999992   0.           0.           0.        ]\n",
      " [  0.83999997  13.68000031   0.96000004   0.           0.           0.        ]\n",
      " [  0.71999997   0.51999998  12.43999958   0.           0.           0.        ]\n",
      " [  0.           0.2          0.          12.55999947   4.           0.16      ]\n",
      " [  0.04         0.           0.           1.03999996  17.44000053   0.        ]\n",
      " [  0.           1.15999997   0.           0.           0.          17.44000053]]\n",
      "Note: training and testing data is not equally distributed amongst classes, \n",
      "so it is normal that more than a 6th of the data is correctly classifier in the last category.\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAzUAAAM+CAYAAAAjKy8vAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11fba9748>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Y_output = model.predict(Xt)\n",
    "Y_pred = np.argmax(Y_output, axis=1)\n",
    "print(np.array(Y_pred))\n",
    "print(\"Accuracy Rate:\")\n",
    "print(metrics.accuracy_score(Yt, Y_pred))\n",
    "print(metrics.confusion_matrix(Yt, Y_pred))\n",
    "\n",
    "print(\"Precision: {}%\".format(100*metrics.precision_score(Yt, Y_pred, average=\"weighted\")))\n",
    "print(\"Recall: {}%\".format(100*metrics.recall_score(Yt, Y_pred, average=\"weighted\")))\n",
    "print(\"f1_score: {}%\".format(100*metrics.f1_score(Yt, Y_pred, average=\"weighted\")))\n",
    "\n",
    "print(\"\")\n",
    "print(\"Confusion Matrix:\")\n",
    "confusion_matrix = metrics.confusion_matrix(Yt, Y_pred)\n",
    "print(confusion_matrix)\n",
    "normalised_confusion_matrix = np.array(confusion_matrix, dtype=np.float32)/np.sum(confusion_matrix)*100\n",
    "\n",
    "print(\"\")\n",
    "print(\"Confusion matrix (normalised to % of total test data):\")\n",
    "print(normalised_confusion_matrix)\n",
    "print(\"Note: training and testing data is not equally distributed amongst classes, \")\n",
    "print(\"so it is normal that more than a 6th of the data is correctly classifier in the last category.\")\n",
    "\n",
    "# Plot Results: \n",
    "width = 12\n",
    "height = 12\n",
    "plt.figure(figsize=(width, height))\n",
    "plt.imshow(\n",
    "    normalised_confusion_matrix, \n",
    "    interpolation='nearest', \n",
    "    cmap=plt.cm.rainbow\n",
    ")\n",
    "plt.title(\"Confusion matrix \\n(normalised to % of total test data)\")\n",
    "plt.colorbar()\n",
    "tick_marks = np.arange(len(LABELS))\n",
    "plt.xticks(tick_marks, LABELS, rotation=90)\n",
    "plt.yticks(tick_marks, LABELS)\n",
    "plt.tight_layout()\n",
    "plt.ylabel('True label')\n",
    "plt.xlabel('Predicted label')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}