Chapter03/Activity3.02/Activity3_02.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Activity 3.02\n"
   ]
  },
  {
   "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",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from tensorflow import random"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt \n",
    "import matplotlib\n",
    "%matplotlib inline "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of Examples in the Dataset =  1385\n",
      "Number of Features for each example =  28\n",
      "Possible Output Classes =  [0 1]\n"
     ]
    }
   ],
   "source": [
    "# Load The dataset\n",
    "X = pd.read_csv('../data/HCV_feats.csv')\n",
    "y = pd.read_csv('../data/HCV_target.csv')\n",
    "\n",
    "# Print the sizes of the dataset\n",
    "print(\"Number of Examples in the Dataset = \", X.shape[0])\n",
    "print(\"Number of Features for each example = \", X.shape[1]) \n",
    "print(\"Possible Output Classes = \", y['AdvancedFibrosis'].unique())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Set up a seed for random number generator so the result will be reproducible\n",
    "\n",
    "Split the dataset into training set and test set with a 80-20 ratio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1108, 28)\n",
      "(1108, 1)\n",
      "(277, 28)\n",
      "(277, 1)\n",
      "Number of examples in training set =  1108\n",
      "Number of examples in test set =  277\n"
     ]
    }
   ],
   "source": [
    "seed = 1\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "sc = StandardScaler()\n",
    "X = pd.DataFrame(sc.fit_transform(X), columns=X.columns)\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=seed)\n",
    "\n",
    "# Print the information regarding dataset sizes\n",
    "print(X_train.shape)\n",
    "print(y_train.shape)\n",
    "print(X_test.shape)\n",
    "print(y_test.shape)\n",
    "print (\"Number of examples in training set = \", X_train.shape[0])\n",
    "print (\"Number of examples in test set = \", X_test.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 3)                 87        \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 1)                 4         \n",
      "=================================================================\n",
      "Total params: 91\n",
      "Trainable params: 91\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "# define the keras model\n",
    "classifier = Sequential()\n",
    "classifier.add(Dense(units = 3, activation = 'tanh', input_dim=X_train.shape[1]))\n",
    "classifier.add(Dense(units = 1, activation = 'sigmoid'))\n",
    "classifier.compile(optimizer = 'sgd', loss = 'binary_crossentropy', metrics = ['accuracy'])\n",
    "\n",
    "classifier.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 997 samples, validate on 111 samples\n",
      "Epoch 1/100\n",
      "997/997 [==============================] - 0s 179us/step - loss: 0.7972 - accuracy: 0.4714 - val_loss: 0.7460 - val_accuracy: 0.5676\n",
      "Epoch 2/100\n",
      "997/997 [==============================] - 0s 56us/step - loss: 0.7795 - accuracy: 0.4764 - val_loss: 0.7343 - val_accuracy: 0.5676\n",
      "Epoch 3/100\n",
      "997/997 [==============================] - 0s 56us/step - loss: 0.7653 - accuracy: 0.4794 - val_loss: 0.7248 - val_accuracy: 0.5676\n",
      "Epoch 4/100\n",
      "997/997 [==============================] - 0s 64us/step - loss: 0.7536 - accuracy: 0.4875 - val_loss: 0.7169 - val_accuracy: 0.5766\n",
      "Epoch 5/100\n",
      "997/997 [==============================] - 0s 62us/step - loss: 0.7441 - accuracy: 0.4945 - val_loss: 0.7104 - val_accuracy: 0.5766\n",
      "Epoch 6/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.7363 - accuracy: 0.5035 - val_loss: 0.7050 - val_accuracy: 0.5676\n",
      "Epoch 7/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.7297 - accuracy: 0.5075 - val_loss: 0.7005 - val_accuracy: 0.5586\n",
      "Epoch 8/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.7243 - accuracy: 0.5115 - val_loss: 0.6967 - val_accuracy: 0.5586\n",
      "Epoch 9/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.7197 - accuracy: 0.5206 - val_loss: 0.6935 - val_accuracy: 0.5495\n",
      "Epoch 10/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.7158 - accuracy: 0.5236 - val_loss: 0.6908 - val_accuracy: 0.5495\n",
      "Epoch 11/100\n",
      "997/997 [==============================] - 0s 61us/step - loss: 0.7125 - accuracy: 0.5256 - val_loss: 0.6886 - val_accuracy: 0.5766\n",
      "Epoch 12/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.7097 - accuracy: 0.5266 - val_loss: 0.6866 - val_accuracy: 0.5766\n",
      "Epoch 13/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.7072 - accuracy: 0.5246 - val_loss: 0.6849 - val_accuracy: 0.5766\n",
      "Epoch 14/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.7051 - accuracy: 0.5276 - val_loss: 0.6835 - val_accuracy: 0.5766\n",
      "Epoch 15/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.7032 - accuracy: 0.5216 - val_loss: 0.6823 - val_accuracy: 0.5766\n",
      "Epoch 16/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.7016 - accuracy: 0.5236 - val_loss: 0.6812 - val_accuracy: 0.5766\n",
      "Epoch 17/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.7002 - accuracy: 0.5236 - val_loss: 0.6803 - val_accuracy: 0.5766\n",
      "Epoch 18/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6990 - accuracy: 0.5236 - val_loss: 0.6795 - val_accuracy: 0.5676\n",
      "Epoch 19/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6979 - accuracy: 0.5246 - val_loss: 0.6788 - val_accuracy: 0.5676\n",
      "Epoch 20/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6969 - accuracy: 0.5266 - val_loss: 0.6782 - val_accuracy: 0.5766\n",
      "Epoch 21/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6960 - accuracy: 0.5266 - val_loss: 0.6777 - val_accuracy: 0.5856\n",
      "Epoch 22/100\n",
      "997/997 [==============================] - 0s 64us/step - loss: 0.6952 - accuracy: 0.5256 - val_loss: 0.6772 - val_accuracy: 0.5856\n",
      "Epoch 23/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6945 - accuracy: 0.5266 - val_loss: 0.6768 - val_accuracy: 0.5766\n",
      "Epoch 24/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6938 - accuracy: 0.5266 - val_loss: 0.6765 - val_accuracy: 0.5766\n",
      "Epoch 25/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6932 - accuracy: 0.5256 - val_loss: 0.6762 - val_accuracy: 0.5766\n",
      "Epoch 26/100\n",
      "997/997 [==============================] - 0s 57us/step - loss: 0.6927 - accuracy: 0.5256 - val_loss: 0.6759 - val_accuracy: 0.5856\n",
      "Epoch 27/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6922 - accuracy: 0.5276 - val_loss: 0.6756 - val_accuracy: 0.5856\n",
      "Epoch 28/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6918 - accuracy: 0.5306 - val_loss: 0.6754 - val_accuracy: 0.5856\n",
      "Epoch 29/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6913 - accuracy: 0.5316 - val_loss: 0.6752 - val_accuracy: 0.5856\n",
      "Epoch 30/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6910 - accuracy: 0.5306 - val_loss: 0.6750 - val_accuracy: 0.5856\n",
      "Epoch 31/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6906 - accuracy: 0.5296 - val_loss: 0.6749 - val_accuracy: 0.5856\n",
      "Epoch 32/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6903 - accuracy: 0.5326 - val_loss: 0.6747 - val_accuracy: 0.5856\n",
      "Epoch 33/100\n",
      "997/997 [==============================] - 0s 60us/step - loss: 0.6899 - accuracy: 0.5336 - val_loss: 0.6746 - val_accuracy: 0.5946\n",
      "Epoch 34/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6896 - accuracy: 0.5386 - val_loss: 0.6745 - val_accuracy: 0.5946\n",
      "Epoch 35/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6894 - accuracy: 0.5396 - val_loss: 0.6744 - val_accuracy: 0.6036\n",
      "Epoch 36/100\n",
      "997/997 [==============================] - 0s 64us/step - loss: 0.6891 - accuracy: 0.5406 - val_loss: 0.6743 - val_accuracy: 0.6126\n",
      "Epoch 37/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6889 - accuracy: 0.5426 - val_loss: 0.6742 - val_accuracy: 0.6036\n",
      "Epoch 38/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6886 - accuracy: 0.5426 - val_loss: 0.6741 - val_accuracy: 0.6036\n",
      "Epoch 39/100\n",
      "997/997 [==============================] - 0s 61us/step - loss: 0.6884 - accuracy: 0.5436 - val_loss: 0.6740 - val_accuracy: 0.6036\n",
      "Epoch 40/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6882 - accuracy: 0.5486 - val_loss: 0.6739 - val_accuracy: 0.6036\n",
      "Epoch 41/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6879 - accuracy: 0.5507 - val_loss: 0.6738 - val_accuracy: 0.6036\n",
      "Epoch 42/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.6877 - accuracy: 0.5527 - val_loss: 0.6738 - val_accuracy: 0.6036\n",
      "Epoch 43/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6875 - accuracy: 0.5527 - val_loss: 0.6737 - val_accuracy: 0.5946\n",
      "Epoch 44/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6873 - accuracy: 0.5527 - val_loss: 0.6736 - val_accuracy: 0.6036\n",
      "Epoch 45/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6871 - accuracy: 0.5537 - val_loss: 0.6736 - val_accuracy: 0.6126\n",
      "Epoch 46/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6870 - accuracy: 0.5547 - val_loss: 0.6735 - val_accuracy: 0.6036\n",
      "Epoch 47/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.6868 - accuracy: 0.5547 - val_loss: 0.6734 - val_accuracy: 0.6036\n",
      "Epoch 48/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6866 - accuracy: 0.5537 - val_loss: 0.6734 - val_accuracy: 0.6036\n",
      "Epoch 49/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6864 - accuracy: 0.5567 - val_loss: 0.6733 - val_accuracy: 0.6036\n",
      "Epoch 50/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6862 - accuracy: 0.5597 - val_loss: 0.6733 - val_accuracy: 0.6036\n",
      "Epoch 51/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.6861 - accuracy: 0.5627 - val_loss: 0.6732 - val_accuracy: 0.6036\n",
      "Epoch 52/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6859 - accuracy: 0.5617 - val_loss: 0.6732 - val_accuracy: 0.6036\n",
      "Epoch 53/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6857 - accuracy: 0.5617 - val_loss: 0.6731 - val_accuracy: 0.6036\n",
      "Epoch 54/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6855 - accuracy: 0.5617 - val_loss: 0.6731 - val_accuracy: 0.6036\n",
      "Epoch 55/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.6854 - accuracy: 0.5617 - val_loss: 0.6730 - val_accuracy: 0.6036\n",
      "Epoch 56/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6852 - accuracy: 0.5627 - val_loss: 0.6730 - val_accuracy: 0.5946\n",
      "Epoch 57/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6850 - accuracy: 0.5637 - val_loss: 0.6729 - val_accuracy: 0.5946\n",
      "Epoch 58/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.6849 - accuracy: 0.5607 - val_loss: 0.6729 - val_accuracy: 0.5946\n",
      "Epoch 59/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.6847 - accuracy: 0.5627 - val_loss: 0.6728 - val_accuracy: 0.5946\n",
      "Epoch 60/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6845 - accuracy: 0.5607 - val_loss: 0.6728 - val_accuracy: 0.5946\n",
      "Epoch 61/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6844 - accuracy: 0.5597 - val_loss: 0.6728 - val_accuracy: 0.5946\n",
      "Epoch 62/100\n",
      "997/997 [==============================] - 0s 46us/step - loss: 0.6842 - accuracy: 0.5607 - val_loss: 0.6727 - val_accuracy: 0.5946\n",
      "Epoch 63/100\n",
      "997/997 [==============================] - 0s 46us/step - loss: 0.6840 - accuracy: 0.5617 - val_loss: 0.6727 - val_accuracy: 0.5946\n",
      "Epoch 64/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6839 - accuracy: 0.5607 - val_loss: 0.6726 - val_accuracy: 0.5946\n",
      "Epoch 65/100\n",
      "997/997 [==============================] - 0s 44us/step - loss: 0.6837 - accuracy: 0.5597 - val_loss: 0.6726 - val_accuracy: 0.5946\n",
      "Epoch 66/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.6835 - accuracy: 0.5607 - val_loss: 0.6726 - val_accuracy: 0.5946\n",
      "Epoch 67/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6834 - accuracy: 0.5617 - val_loss: 0.6725 - val_accuracy: 0.5946\n",
      "Epoch 68/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6832 - accuracy: 0.5617 - val_loss: 0.6725 - val_accuracy: 0.5946\n",
      "Epoch 69/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6830 - accuracy: 0.5607 - val_loss: 0.6725 - val_accuracy: 0.5856\n",
      "Epoch 70/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6829 - accuracy: 0.5597 - val_loss: 0.6724 - val_accuracy: 0.5856\n",
      "Epoch 71/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.6827 - accuracy: 0.5607 - val_loss: 0.6724 - val_accuracy: 0.5856\n",
      "Epoch 72/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6825 - accuracy: 0.5607 - val_loss: 0.6724 - val_accuracy: 0.5856\n",
      "Epoch 73/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6823 - accuracy: 0.5617 - val_loss: 0.6724 - val_accuracy: 0.5856\n",
      "Epoch 74/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6822 - accuracy: 0.5617 - val_loss: 0.6724 - val_accuracy: 0.5856\n",
      "Epoch 75/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6820 - accuracy: 0.5627 - val_loss: 0.6723 - val_accuracy: 0.5856\n",
      "Epoch 76/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6818 - accuracy: 0.5637 - val_loss: 0.6723 - val_accuracy: 0.5856\n",
      "Epoch 77/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.6817 - accuracy: 0.5667 - val_loss: 0.6723 - val_accuracy: 0.5856\n",
      "Epoch 78/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.6815 - accuracy: 0.5697 - val_loss: 0.6723 - val_accuracy: 0.5856\n",
      "Epoch 79/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6813 - accuracy: 0.5707 - val_loss: 0.6723 - val_accuracy: 0.5946\n",
      "Epoch 80/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6811 - accuracy: 0.5717 - val_loss: 0.6723 - val_accuracy: 0.5946\n",
      "Epoch 81/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6810 - accuracy: 0.5717 - val_loss: 0.6723 - val_accuracy: 0.5946\n",
      "Epoch 82/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6808 - accuracy: 0.5707 - val_loss: 0.6723 - val_accuracy: 0.5946\n",
      "Epoch 83/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6806 - accuracy: 0.5727 - val_loss: 0.6723 - val_accuracy: 0.5856\n",
      "Epoch 84/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.6805 - accuracy: 0.5737 - val_loss: 0.6723 - val_accuracy: 0.5856\n",
      "Epoch 85/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.6803 - accuracy: 0.5747 - val_loss: 0.6723 - val_accuracy: 0.5856\n",
      "Epoch 86/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6801 - accuracy: 0.5727 - val_loss: 0.6723 - val_accuracy: 0.5766\n",
      "Epoch 87/100\n",
      "997/997 [==============================] - 0s 60us/step - loss: 0.6799 - accuracy: 0.5707 - val_loss: 0.6723 - val_accuracy: 0.5766\n",
      "Epoch 88/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6798 - accuracy: 0.5707 - val_loss: 0.6723 - val_accuracy: 0.5766\n",
      "Epoch 89/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6796 - accuracy: 0.5677 - val_loss: 0.6723 - val_accuracy: 0.5766\n",
      "Epoch 90/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6794 - accuracy: 0.5687 - val_loss: 0.6724 - val_accuracy: 0.5766\n",
      "Epoch 91/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6792 - accuracy: 0.5707 - val_loss: 0.6724 - val_accuracy: 0.5856\n",
      "Epoch 92/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6791 - accuracy: 0.5707 - val_loss: 0.6724 - val_accuracy: 0.5856\n",
      "Epoch 93/100\n",
      "997/997 [==============================] - 0s 60us/step - loss: 0.6789 - accuracy: 0.5707 - val_loss: 0.6724 - val_accuracy: 0.5856\n",
      "Epoch 94/100\n",
      "997/997 [==============================] - 0s 57us/step - loss: 0.6787 - accuracy: 0.5707 - val_loss: 0.6725 - val_accuracy: 0.5946\n",
      "Epoch 95/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6786 - accuracy: 0.5697 - val_loss: 0.6725 - val_accuracy: 0.5946\n",
      "Epoch 96/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6784 - accuracy: 0.5697 - val_loss: 0.6725 - val_accuracy: 0.5946\n",
      "Epoch 97/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6782 - accuracy: 0.5697 - val_loss: 0.6725 - val_accuracy: 0.5946\n",
      "Epoch 98/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6780 - accuracy: 0.5687 - val_loss: 0.6726 - val_accuracy: 0.5856\n",
      "Epoch 99/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6779 - accuracy: 0.5657 - val_loss: 0.6726 - val_accuracy: 0.5766\n",
      "Epoch 100/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6777 - accuracy: 0.5667 - val_loss: 0.6726 - val_accuracy: 0.5766\n"
     ]
    }
   ],
   "source": [
    "# train the model while storing all loss values\n",
    "history=classifier.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_split=0.1, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x13a57c160>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0) \n",
    "\n",
    "# plot training error and test error plots \n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on training set = 57.472%\n",
      "Best Accuracy on validation set = 61.261%\n",
      "277/277 [==============================] - 0s 36us/step\n",
      "The loss on the test set is 0.7101 and the accuracy is 49.458%\n"
     ]
    }
   ],
   "source": [
    "# print the best accuracy reached on training set and the test set\n",
    "print(f\"Best Accuracy on training set = {max(history.history['accuracy'])*100:.3f}%\")\n",
    "print(f\"Best Accuracy on validation set = {max(history.history['val_accuracy'])*100:.3f}%\") \n",
    "\n",
    "test_loss, test_acc = classifier.evaluate(X_test, y_test['AdvancedFibrosis'])\n",
    "print(f'The loss on the test set is {test_loss:.4f} and the accuracy is {test_acc*100:.3f}%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_3 (Dense)              (None, 4)                 116       \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 2)                 10        \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 1)                 3         \n",
      "=================================================================\n",
      "Total params: 129\n",
      "Trainable params: 129\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# set up a seed for random number generator so the result will be reproducible\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "# define the keras model\n",
    "classifier = Sequential()\n",
    "classifier.add(Dense(units = 4, activation = 'tanh', input_dim = X_train.shape[1]))\n",
    "classifier.add(Dense(units = 2, activation = 'tanh'))\n",
    "classifier.add(Dense(units = 1, activation = 'sigmoid'))\n",
    "classifier.compile(optimizer = 'sgd', loss = 'binary_crossentropy', metrics = ['accuracy'])\n",
    "\n",
    "classifier.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 997 samples, validate on 111 samples\n",
      "Epoch 1/100\n",
      "997/997 [==============================] - 0s 199us/step - loss: 0.7468 - accuracy: 0.4945 - val_loss: 0.7153 - val_accuracy: 0.5676\n",
      "Epoch 2/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.7376 - accuracy: 0.4935 - val_loss: 0.7101 - val_accuracy: 0.5586\n",
      "Epoch 3/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.7302 - accuracy: 0.4945 - val_loss: 0.7060 - val_accuracy: 0.5586\n",
      "Epoch 4/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.7243 - accuracy: 0.5035 - val_loss: 0.7028 - val_accuracy: 0.5586\n",
      "Epoch 5/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.7195 - accuracy: 0.5035 - val_loss: 0.7002 - val_accuracy: 0.5586\n",
      "Epoch 6/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.7155 - accuracy: 0.5015 - val_loss: 0.6981 - val_accuracy: 0.5405\n",
      "Epoch 7/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.7123 - accuracy: 0.4995 - val_loss: 0.6964 - val_accuracy: 0.5225\n",
      "Epoch 8/100\n",
      "997/997 [==============================] - 0s 46us/step - loss: 0.7096 - accuracy: 0.5035 - val_loss: 0.6951 - val_accuracy: 0.5315\n",
      "Epoch 9/100\n",
      "997/997 [==============================] - 0s 45us/step - loss: 0.7073 - accuracy: 0.5075 - val_loss: 0.6940 - val_accuracy: 0.5315\n",
      "Epoch 10/100\n",
      "997/997 [==============================] - 0s 45us/step - loss: 0.7054 - accuracy: 0.5055 - val_loss: 0.6931 - val_accuracy: 0.5315\n",
      "Epoch 11/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.7038 - accuracy: 0.5045 - val_loss: 0.6924 - val_accuracy: 0.5315\n",
      "Epoch 12/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.7025 - accuracy: 0.4985 - val_loss: 0.6918 - val_accuracy: 0.5225\n",
      "Epoch 13/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.7013 - accuracy: 0.5055 - val_loss: 0.6914 - val_accuracy: 0.5315\n",
      "Epoch 14/100\n",
      "997/997 [==============================] - 0s 46us/step - loss: 0.7003 - accuracy: 0.5055 - val_loss: 0.6910 - val_accuracy: 0.5315\n",
      "Epoch 15/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.6994 - accuracy: 0.5035 - val_loss: 0.6907 - val_accuracy: 0.5315\n",
      "Epoch 16/100\n",
      "997/997 [==============================] - 0s 45us/step - loss: 0.6987 - accuracy: 0.5055 - val_loss: 0.6905 - val_accuracy: 0.5586\n",
      "Epoch 17/100\n",
      "997/997 [==============================] - 0s 44us/step - loss: 0.6981 - accuracy: 0.5035 - val_loss: 0.6903 - val_accuracy: 0.5586\n",
      "Epoch 18/100\n",
      "997/997 [==============================] - 0s 46us/step - loss: 0.6975 - accuracy: 0.5025 - val_loss: 0.6901 - val_accuracy: 0.5586\n",
      "Epoch 19/100\n",
      "997/997 [==============================] - 0s 47us/step - loss: 0.6970 - accuracy: 0.5005 - val_loss: 0.6900 - val_accuracy: 0.5405\n",
      "Epoch 20/100\n",
      "997/997 [==============================] - 0s 48us/step - loss: 0.6966 - accuracy: 0.5025 - val_loss: 0.6899 - val_accuracy: 0.5405\n",
      "Epoch 21/100\n",
      "997/997 [==============================] - 0s 46us/step - loss: 0.6962 - accuracy: 0.5035 - val_loss: 0.6898 - val_accuracy: 0.5225\n",
      "Epoch 22/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6958 - accuracy: 0.5025 - val_loss: 0.6898 - val_accuracy: 0.5225\n",
      "Epoch 23/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6955 - accuracy: 0.5015 - val_loss: 0.6897 - val_accuracy: 0.5225\n",
      "Epoch 24/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6953 - accuracy: 0.5015 - val_loss: 0.6897 - val_accuracy: 0.5135\n",
      "Epoch 25/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6950 - accuracy: 0.5005 - val_loss: 0.6896 - val_accuracy: 0.5135\n",
      "Epoch 26/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6948 - accuracy: 0.4955 - val_loss: 0.6896 - val_accuracy: 0.4955\n",
      "Epoch 27/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6946 - accuracy: 0.5005 - val_loss: 0.6896 - val_accuracy: 0.4955\n",
      "Epoch 28/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6944 - accuracy: 0.5015 - val_loss: 0.6896 - val_accuracy: 0.5045\n",
      "Epoch 29/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6942 - accuracy: 0.5055 - val_loss: 0.6896 - val_accuracy: 0.5045\n",
      "Epoch 30/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6941 - accuracy: 0.5075 - val_loss: 0.6896 - val_accuracy: 0.5045\n",
      "Epoch 31/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6939 - accuracy: 0.5075 - val_loss: 0.6896 - val_accuracy: 0.4955\n",
      "Epoch 32/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6938 - accuracy: 0.5065 - val_loss: 0.6896 - val_accuracy: 0.4955\n",
      "Epoch 33/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6936 - accuracy: 0.5085 - val_loss: 0.6896 - val_accuracy: 0.5045\n",
      "Epoch 34/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6935 - accuracy: 0.5115 - val_loss: 0.6896 - val_accuracy: 0.5135\n",
      "Epoch 35/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6934 - accuracy: 0.5125 - val_loss: 0.6896 - val_accuracy: 0.5135\n",
      "Epoch 36/100\n",
      "997/997 [==============================] - 0s 56us/step - loss: 0.6933 - accuracy: 0.5115 - val_loss: 0.6896 - val_accuracy: 0.5135\n",
      "Epoch 37/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6932 - accuracy: 0.5115 - val_loss: 0.6896 - val_accuracy: 0.5135\n",
      "Epoch 38/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6931 - accuracy: 0.5115 - val_loss: 0.6895 - val_accuracy: 0.5225\n",
      "Epoch 39/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6930 - accuracy: 0.5125 - val_loss: 0.6895 - val_accuracy: 0.5135\n",
      "Epoch 40/100\n",
      "997/997 [==============================] - 0s 60us/step - loss: 0.6929 - accuracy: 0.5145 - val_loss: 0.6895 - val_accuracy: 0.5135\n",
      "Epoch 41/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6928 - accuracy: 0.5115 - val_loss: 0.6895 - val_accuracy: 0.5135\n",
      "Epoch 42/100\n",
      "997/997 [==============================] - 0s 61us/step - loss: 0.6927 - accuracy: 0.5135 - val_loss: 0.6895 - val_accuracy: 0.5135\n",
      "Epoch 43/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6926 - accuracy: 0.5196 - val_loss: 0.6895 - val_accuracy: 0.5135\n",
      "Epoch 44/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6926 - accuracy: 0.5206 - val_loss: 0.6895 - val_accuracy: 0.5135\n",
      "Epoch 45/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6925 - accuracy: 0.5216 - val_loss: 0.6895 - val_accuracy: 0.5045\n",
      "Epoch 46/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6924 - accuracy: 0.5246 - val_loss: 0.6895 - val_accuracy: 0.5045\n",
      "Epoch 47/100\n",
      "997/997 [==============================] - 0s 105us/step - loss: 0.6923 - accuracy: 0.5256 - val_loss: 0.6895 - val_accuracy: 0.5045\n",
      "Epoch 48/100\n",
      "997/997 [==============================] - 0s 59us/step - loss: 0.6923 - accuracy: 0.5266 - val_loss: 0.6895 - val_accuracy: 0.5225\n",
      "Epoch 49/100\n",
      "997/997 [==============================] - 0s 56us/step - loss: 0.6922 - accuracy: 0.5276 - val_loss: 0.6895 - val_accuracy: 0.5225\n",
      "Epoch 50/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6921 - accuracy: 0.5256 - val_loss: 0.6895 - val_accuracy: 0.5225\n",
      "Epoch 51/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6921 - accuracy: 0.5286 - val_loss: 0.6895 - val_accuracy: 0.5135\n",
      "Epoch 52/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6920 - accuracy: 0.5286 - val_loss: 0.6894 - val_accuracy: 0.5135\n",
      "Epoch 53/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6919 - accuracy: 0.5306 - val_loss: 0.6894 - val_accuracy: 0.5225\n",
      "Epoch 54/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6919 - accuracy: 0.5316 - val_loss: 0.6894 - val_accuracy: 0.5225\n",
      "Epoch 55/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6918 - accuracy: 0.5316 - val_loss: 0.6894 - val_accuracy: 0.5225\n",
      "Epoch 56/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6917 - accuracy: 0.5316 - val_loss: 0.6894 - val_accuracy: 0.5315\n",
      "Epoch 57/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6917 - accuracy: 0.5306 - val_loss: 0.6894 - val_accuracy: 0.5315\n",
      "Epoch 58/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6916 - accuracy: 0.5296 - val_loss: 0.6894 - val_accuracy: 0.5225\n",
      "Epoch 59/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6915 - accuracy: 0.5306 - val_loss: 0.6894 - val_accuracy: 0.5135\n",
      "Epoch 60/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6915 - accuracy: 0.5316 - val_loss: 0.6894 - val_accuracy: 0.5135\n",
      "Epoch 61/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6914 - accuracy: 0.5316 - val_loss: 0.6894 - val_accuracy: 0.5135\n",
      "Epoch 62/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6914 - accuracy: 0.5336 - val_loss: 0.6893 - val_accuracy: 0.5135\n",
      "Epoch 63/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6913 - accuracy: 0.5326 - val_loss: 0.6893 - val_accuracy: 0.5135\n",
      "Epoch 64/100\n",
      "997/997 [==============================] - 0s 54us/step - loss: 0.6912 - accuracy: 0.5366 - val_loss: 0.6893 - val_accuracy: 0.5135\n",
      "Epoch 65/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6912 - accuracy: 0.5376 - val_loss: 0.6893 - val_accuracy: 0.5135\n",
      "Epoch 66/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6911 - accuracy: 0.5366 - val_loss: 0.6893 - val_accuracy: 0.5135\n",
      "Epoch 67/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6910 - accuracy: 0.5356 - val_loss: 0.6893 - val_accuracy: 0.5045\n",
      "Epoch 68/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6910 - accuracy: 0.5386 - val_loss: 0.6893 - val_accuracy: 0.5045\n",
      "Epoch 69/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6909 - accuracy: 0.5396 - val_loss: 0.6893 - val_accuracy: 0.5135\n",
      "Epoch 70/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6908 - accuracy: 0.5416 - val_loss: 0.6893 - val_accuracy: 0.5135\n",
      "Epoch 71/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6908 - accuracy: 0.5416 - val_loss: 0.6893 - val_accuracy: 0.5135\n",
      "Epoch 72/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6907 - accuracy: 0.5426 - val_loss: 0.6892 - val_accuracy: 0.5135\n",
      "Epoch 73/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6906 - accuracy: 0.5416 - val_loss: 0.6892 - val_accuracy: 0.5045\n",
      "Epoch 74/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6906 - accuracy: 0.5426 - val_loss: 0.6892 - val_accuracy: 0.4955\n",
      "Epoch 75/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6905 - accuracy: 0.5436 - val_loss: 0.6892 - val_accuracy: 0.4955\n",
      "Epoch 76/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6904 - accuracy: 0.5456 - val_loss: 0.6892 - val_accuracy: 0.5045\n",
      "Epoch 77/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6904 - accuracy: 0.5446 - val_loss: 0.6892 - val_accuracy: 0.5135\n",
      "Epoch 78/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6903 - accuracy: 0.5456 - val_loss: 0.6892 - val_accuracy: 0.5135\n",
      "Epoch 79/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6902 - accuracy: 0.5466 - val_loss: 0.6892 - val_accuracy: 0.5135\n",
      "Epoch 80/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6902 - accuracy: 0.5476 - val_loss: 0.6892 - val_accuracy: 0.5135\n",
      "Epoch 81/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6901 - accuracy: 0.5476 - val_loss: 0.6892 - val_accuracy: 0.5135\n",
      "Epoch 82/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6900 - accuracy: 0.5496 - val_loss: 0.6892 - val_accuracy: 0.5135\n",
      "Epoch 83/100\n",
      "997/997 [==============================] - 0s 55us/step - loss: 0.6899 - accuracy: 0.5507 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 84/100\n",
      "997/997 [==============================] - 0s 56us/step - loss: 0.6899 - accuracy: 0.5496 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 85/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6898 - accuracy: 0.5496 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 86/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6897 - accuracy: 0.5496 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 87/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6896 - accuracy: 0.5496 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 88/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6896 - accuracy: 0.5517 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 89/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6895 - accuracy: 0.5517 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 90/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6894 - accuracy: 0.5517 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 91/100\n",
      "997/997 [==============================] - 0s 49us/step - loss: 0.6893 - accuracy: 0.5557 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 92/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6892 - accuracy: 0.5557 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 93/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6891 - accuracy: 0.5557 - val_loss: 0.6891 - val_accuracy: 0.5135\n",
      "Epoch 94/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6891 - accuracy: 0.5537 - val_loss: 0.6891 - val_accuracy: 0.5225\n",
      "Epoch 95/100\n",
      "997/997 [==============================] - 0s 53us/step - loss: 0.6890 - accuracy: 0.5547 - val_loss: 0.6891 - val_accuracy: 0.5225\n",
      "Epoch 96/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6889 - accuracy: 0.5547 - val_loss: 0.6890 - val_accuracy: 0.5225\n",
      "Epoch 97/100\n",
      "997/997 [==============================] - 0s 51us/step - loss: 0.6888 - accuracy: 0.5557 - val_loss: 0.6890 - val_accuracy: 0.5225\n",
      "Epoch 98/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6887 - accuracy: 0.5547 - val_loss: 0.6890 - val_accuracy: 0.5225\n",
      "Epoch 99/100\n",
      "997/997 [==============================] - 0s 50us/step - loss: 0.6886 - accuracy: 0.5557 - val_loss: 0.6890 - val_accuracy: 0.5225\n",
      "Epoch 100/100\n",
      "997/997 [==============================] - 0s 52us/step - loss: 0.6885 - accuracy: 0.5557 - val_loss: 0.6890 - val_accuracy: 0.5315\n"
     ]
    }
   ],
   "source": [
    "# train the model while storing all loss values\n",
    "history=classifier.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_split=0.1, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x13ab84400>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot training error and test error plots \n",
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on training set = 55.567%\n",
      "Best Accuracy on test set = 56.757%\n",
      "277/277 [==============================] - 0s 22us/step\n",
      "The loss on the test set is 0.6861 and the accuracy is 50.903%\n"
     ]
    }
   ],
   "source": [
    "# print the best accuracy reached on training set and the test set\n",
    "print(f\"Best Accuracy on training set = {max(history.history['accuracy'])*100:.3f}%\")\n",
    "print(f\"Best Accuracy on test set = {max(history.history['val_accuracy'])*100:.3f}%\")\n",
    "\n",
    "test_loss, test_acc = classifier.evaluate(X_test, y_test['AdvancedFibrosis'])\n",
    "print(f'The loss on the test set is {test_loss:.4f} and the accuracy is {test_acc*100:.3f}%')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "py3.7",
   "language": "python",
   "name": "py3.7"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}