heart-disease-prediction-with-neural-networks.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Heart Disease Prediction with Neural Networks**\n",
    "\n",
    "I chose this to be my very first Kernel on Kaggle, as in fact I quite enjoy the applications of Machine learning and how it impacts our lives. I am an *absolute beginner* as well and just breaking into the field. Thus, this kernel is more focused for learning purposes and exploration. Please feel free to give advice, recommendations/ better approaches or whatsover on the code below.\n",
    "\n",
    "The heart disease data-set is a quite small dataset with today's modern standards. In this kernel first, I'll do a small bit of data exploration then, I try to create a very simple and straightforward model. That we analyze later, and try to improve its performance.\n",
    "\n",
    "First of all, lets add all packages that we will use."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import packages that we will be working with.\n",
    "import os\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from keras.layers import Dense\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "from tensorflow.python.keras.models import Sequential\n",
    "from tensorflow.python.keras.layers import Dense\n",
    "from tensorflow.python.keras.optimizers import Adam\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report, accuracy_score\n",
    "\n",
    "#print(os.listdir(\"../input\"))\n",
    "np.random.seed(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Dataset**\n",
    "\n",
    "Lets load the dataset, and check its components. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   age  sex  cp  trestbps  chol  fbs  restecg  thalach  exang  oldpeak  slope  \\\n",
      "0   63    1   3       145   233    1        0      150      0      2.3      0   \n",
      "1   37    1   2       130   250    0        1      187      0      3.5      0   \n",
      "2   41    0   1       130   204    0        0      172      0      1.4      2   \n",
      "\n",
      "   ca  thal  target  \n",
      "0   0     1       1  \n",
      "1   0     2       1  \n",
      "2   0     2       1  \n",
      "age           int64\n",
      "sex           int64\n",
      "cp            int64\n",
      "trestbps      int64\n",
      "chol          int64\n",
      "fbs           int64\n",
      "restecg       int64\n",
      "thalach       int64\n",
      "exang         int64\n",
      "oldpeak     float64\n",
      "slope         int64\n",
      "ca            int64\n",
      "thal          int64\n",
      "target        int64\n",
      "dtype: object\n"
     ]
    }
   ],
   "source": [
    "# Load the dataset, and view couple of the first rows.\n",
    "data = pd.read_csv(\"heart.csv\")\n",
    "print(data.head(3))\n",
    "\n",
    "# Check the datatypes\n",
    "print(data.dtypes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So far we have a dataframe that containts all of our data. However we need to split to training data and target labels for learning. Then split the dataset into a train/test sets.\n",
    "\n",
    "We do that in the following block of code."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Features Size: (24, 13)\n",
      "Test Features Size: (61, 13)\n",
      "Train Labels Size: (24,)\n",
      "Test Labels Size: (61,)\n"
     ]
    }
   ],
   "source": [
    "# At this moment we have a dataframe that contains all of the heart.csv data. However we need to\n",
    "# Separate them to [X, Y]. Where our target labels are 'Y', and 'X' is our training data.\n",
    "Y = data.target.values\n",
    "X = data.drop(['target'], axis=1)\n",
    "\n",
    "# Now split to train/test with 80% training data, and 20% test data.\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.2)\n",
    "X_train, X_train2, Y_train, Y_train2 = train_test_split(X_train, Y_train, test_size=0.9)\n",
    "\n",
    "# Check dimensions of both sets.\n",
    "print(\"Train Features Size:\", X_train.shape)\n",
    "print(\"Test Features Size:\", X_test.shape)\n",
    "print(\"Train Labels Size:\", Y_train.shape)\n",
    "print(\"Test Labels Size:\", Y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Neural Network Model**\n",
    "\n",
    "Lets create a function that we can call later that builds our Neural Network model, and takes in the learning rate as a parameter. The architecture of the Neural Network that we're going to implement is detailed in the below illustration.\n",
    "![](https://raw.githubusercontent.com/3absamad/machine-learning/master/Kaggle/imgs/Kaggle_heart_disease_NN_model.png)\n",
    "In our model, we use `Adam` (Adaptive Momentum) as our optimization algorithm, and set our metrics to `accuracy`. Furthermore, I have used the loss function to be `sparse_categorical_crossentropy` because our traget labels are *integers* and not one hot encoded.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a Neural Network Model\n",
    "\n",
    "def NN_model(learning_rate):\n",
    "    model = Sequential()\n",
    "    model.add(Dense(32, input_dim=13, kernel_initializer='normal', activation='relu'))\n",
    "    model.add(Dense(16, kernel_initializer='normal', activation='relu'))\n",
    "    model.add(Dense(2, activation='softmax'))\n",
    "    Adam(lr=learning_rate)\n",
    "    model.compile(loss='sparse_categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now lets build the NN-model and start training. I chose `learning_rate=0.01`, `epochs=100`, and `batch_size=16`. \n",
    "\n",
    "Training the model for 100 epochs, seems to be pretty fine in order to avoid overfitting. I already performed training with 1000 epochs and around 100 epochs was the reasonable number of epochs for early stopping.\n",
    "\n",
    "Lets take a look at our model summary."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_51 (Dense)             (None, 32)                448       \n",
      "_________________________________________________________________\n",
      "dense_52 (Dense)             (None, 16)                528       \n",
      "_________________________________________________________________\n",
      "dense_53 (Dense)             (None, 2)                 34        \n",
      "=================================================================\n",
      "Total params: 1,010\n",
      "Trainable params: 1,010\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "# Build a NN-model, and start training\n",
    "learning_rate = 0.01\n",
    "model = NN_model(learning_rate)\n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we train the network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "_kg_hide-output": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 24 samples, validate on 61 samples\n",
      "Epoch 1/100\n",
      " - 1s - loss: 1.1109 - acc: 0.5000 - val_loss: 0.7646 - val_acc: 0.5082\n",
      "Epoch 2/100\n",
      " - 0s - loss: 0.6487 - acc: 0.6250 - val_loss: 0.6957 - val_acc: 0.5902\n",
      "Epoch 3/100\n",
      " - 0s - loss: 0.7790 - acc: 0.5833 - val_loss: 0.7723 - val_acc: 0.5738\n",
      "Epoch 4/100\n",
      " - 0s - loss: 0.7816 - acc: 0.5000 - val_loss: 0.6581 - val_acc: 0.6066\n",
      "Epoch 5/100\n",
      " - 0s - loss: 0.6575 - acc: 0.6250 - val_loss: 0.6313 - val_acc: 0.6230\n",
      "Epoch 6/100\n",
      " - 0s - loss: 0.5716 - acc: 0.7500 - val_loss: 0.7035 - val_acc: 0.4918\n",
      "Epoch 7/100\n",
      " - 0s - loss: 0.6603 - acc: 0.5833 - val_loss: 0.7716 - val_acc: 0.4918\n",
      "Epoch 8/100\n",
      " - 0s - loss: 0.6940 - acc: 0.5417 - val_loss: 0.7308 - val_acc: 0.5082\n",
      "Epoch 9/100\n",
      " - 0s - loss: 0.6556 - acc: 0.5833 - val_loss: 0.6558 - val_acc: 0.6557\n",
      "Epoch 10/100\n",
      " - 0s - loss: 0.6158 - acc: 0.6667 - val_loss: 0.6219 - val_acc: 0.6557\n",
      "Epoch 11/100\n",
      " - 0s - loss: 0.5907 - acc: 0.6250 - val_loss: 0.6210 - val_acc: 0.6885\n",
      "Epoch 12/100\n",
      " - 0s - loss: 0.6089 - acc: 0.6250 - val_loss: 0.6266 - val_acc: 0.6721\n",
      "Epoch 13/100\n",
      " - 0s - loss: 0.6111 - acc: 0.6250 - val_loss: 0.6209 - val_acc: 0.6557\n",
      "Epoch 14/100\n",
      " - 0s - loss: 0.6008 - acc: 0.6250 - val_loss: 0.6191 - val_acc: 0.6393\n",
      "Epoch 15/100\n",
      " - 0s - loss: 0.5735 - acc: 0.6250 - val_loss: 0.6268 - val_acc: 0.6557\n",
      "Epoch 16/100\n",
      " - 0s - loss: 0.5668 - acc: 0.7917 - val_loss: 0.6336 - val_acc: 0.6393\n",
      "Epoch 17/100\n",
      " - 0s - loss: 0.5670 - acc: 0.7500 - val_loss: 0.6362 - val_acc: 0.6393\n",
      "Epoch 18/100\n",
      " - 0s - loss: 0.5638 - acc: 0.7917 - val_loss: 0.6298 - val_acc: 0.6393\n",
      "Epoch 19/100\n",
      " - 0s - loss: 0.5549 - acc: 0.7917 - val_loss: 0.6258 - val_acc: 0.6230\n",
      "Epoch 20/100\n",
      " - 0s - loss: 0.5511 - acc: 0.7917 - val_loss: 0.6206 - val_acc: 0.6557\n",
      "Epoch 21/100\n",
      " - 0s - loss: 0.5516 - acc: 0.7500 - val_loss: 0.6194 - val_acc: 0.6393\n",
      "Epoch 22/100\n",
      " - 0s - loss: 0.5522 - acc: 0.6667 - val_loss: 0.6183 - val_acc: 0.6393\n",
      "Epoch 23/100\n",
      " - 0s - loss: 0.5472 - acc: 0.6250 - val_loss: 0.6177 - val_acc: 0.6393\n",
      "Epoch 24/100\n",
      " - 0s - loss: 0.5436 - acc: 0.6667 - val_loss: 0.6179 - val_acc: 0.6557\n",
      "Epoch 25/100\n",
      " - 0s - loss: 0.5402 - acc: 0.7083 - val_loss: 0.6187 - val_acc: 0.6557\n",
      "Epoch 26/100\n",
      " - 0s - loss: 0.5411 - acc: 0.7083 - val_loss: 0.6208 - val_acc: 0.6230\n",
      "Epoch 27/100\n",
      " - 0s - loss: 0.5307 - acc: 0.8333 - val_loss: 0.6311 - val_acc: 0.6557\n",
      "Epoch 28/100\n",
      " - 0s - loss: 0.5481 - acc: 0.7917 - val_loss: 0.6362 - val_acc: 0.6557\n",
      "Epoch 29/100\n",
      " - 0s - loss: 0.5346 - acc: 0.7917 - val_loss: 0.6229 - val_acc: 0.6230\n",
      "Epoch 30/100\n",
      " - 0s - loss: 0.5225 - acc: 0.8333 - val_loss: 0.6185 - val_acc: 0.6393\n",
      "Epoch 31/100\n",
      " - 0s - loss: 0.5225 - acc: 0.7500 - val_loss: 0.6176 - val_acc: 0.6393\n",
      "Epoch 32/100\n",
      " - 0s - loss: 0.5207 - acc: 0.7500 - val_loss: 0.6165 - val_acc: 0.6557\n",
      "Epoch 33/100\n",
      " - 0s - loss: 0.5175 - acc: 0.7500 - val_loss: 0.6190 - val_acc: 0.6393\n",
      "Epoch 34/100\n",
      " - 0s - loss: 0.5191 - acc: 0.7917 - val_loss: 0.6263 - val_acc: 0.6557\n",
      "Epoch 35/100\n",
      " - 0s - loss: 0.5148 - acc: 0.7917 - val_loss: 0.6235 - val_acc: 0.6393\n",
      "Epoch 36/100\n",
      " - 0s - loss: 0.5107 - acc: 0.7917 - val_loss: 0.6189 - val_acc: 0.6557\n",
      "Epoch 37/100\n",
      " - 0s - loss: 0.5087 - acc: 0.7917 - val_loss: 0.6160 - val_acc: 0.6393\n",
      "Epoch 38/100\n",
      " - 0s - loss: 0.5036 - acc: 0.7917 - val_loss: 0.6149 - val_acc: 0.6393\n",
      "Epoch 39/100\n",
      " - 0s - loss: 0.5013 - acc: 0.7917 - val_loss: 0.6152 - val_acc: 0.6557\n",
      "Epoch 40/100\n",
      " - 0s - loss: 0.4973 - acc: 0.7917 - val_loss: 0.6148 - val_acc: 0.6557\n",
      "Epoch 41/100\n",
      " - 0s - loss: 0.4962 - acc: 0.7500 - val_loss: 0.6145 - val_acc: 0.6557\n",
      "Epoch 42/100\n",
      " - 0s - loss: 0.4934 - acc: 0.7917 - val_loss: 0.6161 - val_acc: 0.6230\n",
      "Epoch 43/100\n",
      " - 0s - loss: 0.4878 - acc: 0.8333 - val_loss: 0.6229 - val_acc: 0.6557\n",
      "Epoch 44/100\n",
      " - 0s - loss: 0.4902 - acc: 0.7917 - val_loss: 0.6265 - val_acc: 0.6557\n",
      "Epoch 45/100\n",
      " - 0s - loss: 0.4906 - acc: 0.7917 - val_loss: 0.6231 - val_acc: 0.6557\n",
      "Epoch 46/100\n",
      " - 0s - loss: 0.4916 - acc: 0.7917 - val_loss: 0.6148 - val_acc: 0.6557\n",
      "Epoch 47/100\n",
      " - 0s - loss: 0.4867 - acc: 0.8333 - val_loss: 0.6153 - val_acc: 0.6721\n",
      "Epoch 48/100\n",
      " - 0s - loss: 0.4856 - acc: 0.7500 - val_loss: 0.6157 - val_acc: 0.6721\n",
      "Epoch 49/100\n",
      " - 0s - loss: 0.4811 - acc: 0.7083 - val_loss: 0.6110 - val_acc: 0.6557\n",
      "Epoch 50/100\n",
      " - 0s - loss: 0.4830 - acc: 0.7917 - val_loss: 0.6157 - val_acc: 0.6721\n",
      "Epoch 51/100\n",
      " - 0s - loss: 0.4719 - acc: 0.8333 - val_loss: 0.6144 - val_acc: 0.6885\n",
      "Epoch 52/100\n",
      " - 0s - loss: 0.4782 - acc: 0.8333 - val_loss: 0.6112 - val_acc: 0.6721\n",
      "Epoch 53/100\n",
      " - 0s - loss: 0.4642 - acc: 0.8333 - val_loss: 0.6139 - val_acc: 0.6557\n",
      "Epoch 54/100\n",
      " - 0s - loss: 0.4634 - acc: 0.8333 - val_loss: 0.6128 - val_acc: 0.6557\n",
      "Epoch 55/100\n",
      " - 0s - loss: 0.4606 - acc: 0.8333 - val_loss: 0.6110 - val_acc: 0.6721\n",
      "Epoch 56/100\n",
      " - 0s - loss: 0.4551 - acc: 0.8333 - val_loss: 0.6109 - val_acc: 0.6721\n",
      "Epoch 57/100\n",
      " - 0s - loss: 0.4527 - acc: 0.8333 - val_loss: 0.6111 - val_acc: 0.6721\n",
      "Epoch 58/100\n",
      " - 0s - loss: 0.4577 - acc: 0.7500 - val_loss: 0.6117 - val_acc: 0.6721\n",
      "Epoch 59/100\n",
      " - 0s - loss: 0.4525 - acc: 0.8333 - val_loss: 0.6135 - val_acc: 0.6885\n",
      "Epoch 60/100\n",
      " - 0s - loss: 0.4575 - acc: 0.8333 - val_loss: 0.6197 - val_acc: 0.6557\n",
      "Epoch 61/100\n",
      " - 0s - loss: 0.4504 - acc: 0.8333 - val_loss: 0.6130 - val_acc: 0.6721\n",
      "Epoch 62/100\n",
      " - 0s - loss: 0.4401 - acc: 0.8333 - val_loss: 0.6114 - val_acc: 0.6557\n",
      "Epoch 63/100\n",
      " - 0s - loss: 0.4429 - acc: 0.7500 - val_loss: 0.6197 - val_acc: 0.6721\n",
      "Epoch 64/100\n",
      " - 0s - loss: 0.4481 - acc: 0.7917 - val_loss: 0.6179 - val_acc: 0.6721\n",
      "Epoch 65/100\n",
      " - 0s - loss: 0.4413 - acc: 0.7917 - val_loss: 0.6140 - val_acc: 0.6721\n",
      "Epoch 66/100\n",
      " - 0s - loss: 0.4346 - acc: 0.7917 - val_loss: 0.6185 - val_acc: 0.6885\n",
      "Epoch 67/100\n",
      " - 0s - loss: 0.4348 - acc: 0.8333 - val_loss: 0.6180 - val_acc: 0.6721\n",
      "Epoch 68/100\n",
      " - 0s - loss: 0.4271 - acc: 0.7917 - val_loss: 0.6166 - val_acc: 0.6721\n",
      "Epoch 69/100\n",
      " - 0s - loss: 0.4404 - acc: 0.8333 - val_loss: 0.6189 - val_acc: 0.6721\n",
      "Epoch 70/100\n",
      " - 0s - loss: 0.4500 - acc: 0.7917 - val_loss: 0.6138 - val_acc: 0.6557\n",
      "Epoch 71/100\n",
      " - 0s - loss: 0.4197 - acc: 0.7917 - val_loss: 0.6148 - val_acc: 0.6557\n",
      "Epoch 72/100\n",
      " - 0s - loss: 0.4188 - acc: 0.8750 - val_loss: 0.6151 - val_acc: 0.6557\n",
      "Epoch 73/100\n",
      " - 0s - loss: 0.4148 - acc: 0.8750 - val_loss: 0.6152 - val_acc: 0.6885\n",
      "Epoch 74/100\n",
      " - 0s - loss: 0.4135 - acc: 0.8333 - val_loss: 0.6173 - val_acc: 0.6885\n",
      "Epoch 75/100\n",
      " - 0s - loss: 0.4223 - acc: 0.8333 - val_loss: 0.6177 - val_acc: 0.7049\n",
      "Epoch 76/100\n",
      " - 0s - loss: 0.4133 - acc: 0.8333 - val_loss: 0.6148 - val_acc: 0.6557\n",
      "Epoch 77/100\n",
      " - 0s - loss: 0.4068 - acc: 0.8750 - val_loss: 0.6190 - val_acc: 0.6721\n",
      "Epoch 78/100\n",
      " - 0s - loss: 0.4128 - acc: 0.8750 - val_loss: 0.6201 - val_acc: 0.6721\n",
      "Epoch 79/100\n",
      " - 0s - loss: 0.4013 - acc: 0.8750 - val_loss: 0.6219 - val_acc: 0.6721\n",
      "Epoch 80/100\n",
      " - 0s - loss: 0.4083 - acc: 0.8333 - val_loss: 0.6298 - val_acc: 0.6393\n",
      "Epoch 81/100\n",
      " - 0s - loss: 0.4119 - acc: 0.8333 - val_loss: 0.6234 - val_acc: 0.6721\n",
      "Epoch 82/100\n",
      " - 0s - loss: 0.3895 - acc: 0.8333 - val_loss: 0.6350 - val_acc: 0.6721\n",
      "Epoch 83/100\n",
      " - 0s - loss: 0.4316 - acc: 0.6667 - val_loss: 0.6502 - val_acc: 0.7049\n",
      "Epoch 84/100\n",
      " - 0s - loss: 0.4158 - acc: 0.7083 - val_loss: 0.6245 - val_acc: 0.6885\n",
      "Epoch 85/100\n",
      " - 0s - loss: 0.3861 - acc: 0.8750 - val_loss: 0.6291 - val_acc: 0.6721\n",
      "Epoch 86/100\n",
      " - 0s - loss: 0.4226 - acc: 0.8333 - val_loss: 0.6397 - val_acc: 0.6230\n",
      "Epoch 87/100\n",
      " - 0s - loss: 0.4112 - acc: 0.8333 - val_loss: 0.6218 - val_acc: 0.6885\n",
      "Epoch 88/100\n",
      " - 0s - loss: 0.3837 - acc: 0.8333 - val_loss: 0.6297 - val_acc: 0.6721\n",
      "Epoch 89/100\n",
      " - 0s - loss: 0.3950 - acc: 0.8333 - val_loss: 0.6407 - val_acc: 0.6721\n",
      "Epoch 90/100\n",
      " - 0s - loss: 0.4143 - acc: 0.7500 - val_loss: 0.6276 - val_acc: 0.6721\n",
      "Epoch 91/100\n",
      " - 0s - loss: 0.3684 - acc: 0.8750 - val_loss: 0.6442 - val_acc: 0.6230\n",
      "Epoch 92/100\n",
      " - 0s - loss: 0.4201 - acc: 0.7917 - val_loss: 0.6738 - val_acc: 0.6230\n",
      "Epoch 93/100\n",
      " - 0s - loss: 0.4373 - acc: 0.7917 - val_loss: 0.6376 - val_acc: 0.6721\n",
      "Epoch 94/100\n",
      " - 0s - loss: 0.3841 - acc: 0.8333 - val_loss: 0.6407 - val_acc: 0.6885\n",
      "Epoch 95/100\n",
      " - 0s - loss: 0.3983 - acc: 0.7083 - val_loss: 0.6733 - val_acc: 0.7049\n",
      "Epoch 96/100\n",
      " - 0s - loss: 0.4120 - acc: 0.7083 - val_loss: 0.6550 - val_acc: 0.7049\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 97/100\n",
      " - 0s - loss: 0.3985 - acc: 0.8333 - val_loss: 0.6342 - val_acc: 0.6885\n",
      "Epoch 98/100\n",
      " - 0s - loss: 0.3810 - acc: 0.8750 - val_loss: 0.6315 - val_acc: 0.6885\n",
      "Epoch 99/100\n",
      " - 0s - loss: 0.3667 - acc: 0.8333 - val_loss: 0.6299 - val_acc: 0.6721\n",
      "Epoch 100/100\n",
      " - 0s - loss: 0.3667 - acc: 0.9167 - val_loss: 0.6316 - val_acc: 0.6885\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, Y_train, validation_data=(X_test, Y_test), epochs=100, batch_size=16, verbose=2)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We plot the `Model Accuracy`, and `Model Loss`  vs. the number of `Epochs`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x133d77fd0>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x133a730b8>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the model accuracy vs. number of Epochs\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('Model Accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.legend(['Train', 'Test'])\n",
    "plt.show()\n",
    "\n",
    "# Plot the Loss function vs. number of Epochs\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model Loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.legend(['Train', 'Test'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Evaluating our model's performance, we define the f1 score to be:\n",
    "\n",
    "$$f1 = \\frac{precision^{-1} + recall^{-1}}{2}$$\n",
    "\n",
    "\n",
    "$$precision = \\frac{\\textit{true positives}}{\\textit{true positives} + \\textit{false positives}}$$\n",
    "\n",
    "$$recall = \\frac{\\textit{true positives}}{\\textit{true positives} + \\textit{false negatives}}$$\n",
    "\n",
    "Check this [Wikipedia](http://https://en.wikipedia.org/wiki/Precision_and_recall) page for more details on precision and recall.\n",
    "\n",
    "We compute our model's predictions on the test set `X_test` and print a `classfication_report` from the imported package `sklearn.metrics`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model Accracy: 31.14754098360656 %\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.68      0.86      0.76        35\n",
      "           1       0.71      0.46      0.56        26\n",
      "\n",
      "   micro avg       0.69      0.69      0.69        61\n",
      "   macro avg       0.69      0.66      0.66        61\n",
      "weighted avg       0.69      0.69      0.67        61\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = np.argmax(model.predict(X_test), axis=1)\n",
    "model_accuracy = 100-(accuracy_score(Y_test, predictions)*100)\n",
    "print(\"Model Accracy:\", model_accuracy,\"%\")\n",
    "print(classification_report(Y_test, predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With a model accuracy of around 88% and very similar $f1$ score, we have a very good performing model that generalizes well to the test set."
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}