Credit Card Fraudulent Detection .ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "LABELS = ['Normal','Fraud']\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "from keras.layers import Dense, Input\n",
    "from keras.models import Model, load_model\n",
    "from keras.callbacks import ModelCheckpoint, TensorBoard\n",
    "from sklearn.metrics import confusion_matrix, precision_recall_curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('./creditcardfraud/creditcard.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### EDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#check for null values\n",
    "df.isnull().values.any()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0    284315\n",
      "1       492\n",
      "Name: Class, dtype: int64\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#check for class imbalance \n",
    "count = pd.value_counts(df['Class'])\n",
    "count.plot(kind='bar', rot=0, title='Distribution')\n",
    "plt.xticks(range(2),LABELS)\n",
    "plt.xlabel('Class')\n",
    "plt.ylabel('Number of examples')\n",
    "print(count)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "#check how amount varies with class (normal/fraud)\n",
    "frauds_example = df[df['Class']==1]\n",
    "normal_example = df[df['Class']==0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(492, 31)"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "frauds_example.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(284315, 31)"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "normal_example.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0, 20000)"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#checking variation of amounts with fraud/normal cases\n",
    "f,(sub1,sub2) = plt.subplots(2,1,sharex=True)\n",
    "\n",
    "sub1.hist(normal_example['Amount'],bins=50)\n",
    "sub1.set_title('Normal Cases')\n",
    "\n",
    "sub2.hist(normal_example['Amount'], bins=50)\n",
    "sub2.set_title('Fraud Cases')\n",
    "\n",
    "plt.xlabel('Amount')\n",
    "plt.ylabel('Number of transaction')\n",
    "plt.yscale('log')\n",
    "plt.xlim((0,20000))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Preparing Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Normalize\n",
    "df['Amount'] = StandardScaler().fit_transform(df['Amount'].values.reshape(-1,1))\n",
    "\n",
    "X_train,X_test = train_test_split(df, test_size=0.2, random_state=42)\n",
    "X_train = X_train[X_train.Class == 0]\n",
    "X_train = X_train.drop(['Class'], axis=1)\n",
    "\n",
    "y_test = X_test['Class']\n",
    "X_test = X_test.drop(['Class'], axis=1)\n",
    "\n",
    "X_train = X_train.values\n",
    "X_test  = X_test.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(227451, 29)"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_dim = X_train.shape[1]\n",
    "encoding_dim=7\n",
    "\n",
    "input_layer = Input(shape=(input_dim,), )\n",
    "\n",
    "encoder_1 = Dense(2*encoding_dim, activation=\"tanh\" )(input_layer)\n",
    "encoder_2 = Dense(encoding_dim, activation='relu')(encoder_1)\n",
    "\n",
    "decoder_1 = Dense(encoding_dim, activation='relu')(encoder_2)\n",
    "decoder_2 = Dense(2*encoding_dim, activation='tanh')(decoder_1)\n",
    "out = Dense(input_dim)(decoder_2)\n",
    "\n",
    "auto_encoder = Model(inputs=input_layer, outputs=out)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 227451 samples, validate on 56962 samples\n",
      "Epoch 1/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3000 - acc: 0.6788 - val_loss: 0.3334 - val_acc: 0.6723\n",
      "Epoch 2/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3028 - acc: 0.6783 - val_loss: 0.3317 - val_acc: 0.6831\n",
      "Epoch 3/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.3005 - acc: 0.6801 - val_loss: 0.3365 - val_acc: 0.6831\n",
      "Epoch 4/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2993 - acc: 0.6795 - val_loss: 0.3383 - val_acc: 0.6863\n",
      "Epoch 5/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.3021 - acc: 0.6783 - val_loss: 0.3327 - val_acc: 0.6845\n",
      "Epoch 6/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.3007 - acc: 0.6782 - val_loss: 0.3314 - val_acc: 0.6830\n",
      "Epoch 7/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2997 - acc: 0.6796 - val_loss: 0.3340 - val_acc: 0.6829\n",
      "Epoch 8/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2999 - acc: 0.6804 - val_loss: 0.3329 - val_acc: 0.6845\n",
      "Epoch 9/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3007 - acc: 0.6792 - val_loss: 0.3307 - val_acc: 0.6801\n",
      "Epoch 10/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.3003 - acc: 0.6791 - val_loss: 0.3312 - val_acc: 0.6792\n",
      "Epoch 11/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3002 - acc: 0.6800 - val_loss: 0.3316 - val_acc: 0.6812\n",
      "Epoch 12/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2996 - acc: 0.6797 - val_loss: 0.3326 - val_acc: 0.6886\n",
      "Epoch 13/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3007 - acc: 0.6785 - val_loss: 0.3380 - val_acc: 0.6922\n",
      "Epoch 14/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2999 - acc: 0.6797 - val_loss: 0.3318 - val_acc: 0.6818\n",
      "Epoch 15/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2998 - acc: 0.6795 - val_loss: 0.3336 - val_acc: 0.6794\n",
      "Epoch 16/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2996 - acc: 0.6797 - val_loss: 0.3362 - val_acc: 0.6858\n",
      "Epoch 17/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2992 - acc: 0.6806 - val_loss: 0.3326 - val_acc: 0.6846\n",
      "Epoch 18/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2999 - acc: 0.6801 - val_loss: 0.3303 - val_acc: 0.6793\n",
      "Epoch 19/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2988 - acc: 0.6809 - val_loss: 0.3337 - val_acc: 0.6842\n",
      "Epoch 20/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2997 - acc: 0.6804 - val_loss: 0.3327 - val_acc: 0.6743\n",
      "Epoch 21/500\n",
      "227451/227451 [==============================] - 21s 94us/step - loss: 0.2993 - acc: 0.6815 - val_loss: 0.3325 - val_acc: 0.6793\n",
      "Epoch 22/500\n",
      "227451/227451 [==============================] - 21s 93us/step - loss: 0.2990 - acc: 0.6807 - val_loss: 0.3299 - val_acc: 0.6796\n",
      "Epoch 23/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2992 - acc: 0.6815 - val_loss: 0.3358 - val_acc: 0.6760\n",
      "Epoch 24/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3003 - acc: 0.6808 - val_loss: 0.3407 - val_acc: 0.6827\n",
      "Epoch 25/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2994 - acc: 0.6804 - val_loss: 0.3328 - val_acc: 0.6705\n",
      "Epoch 26/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2986 - acc: 0.6804 - val_loss: 0.3306 - val_acc: 0.6731\n",
      "Epoch 27/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2995 - acc: 0.6806 - val_loss: 0.3325 - val_acc: 0.6846\n",
      "Epoch 28/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3005 - acc: 0.6814 - val_loss: 0.3339 - val_acc: 0.6854\n",
      "Epoch 29/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3004 - acc: 0.6807 - val_loss: 0.3502 - val_acc: 0.6662\n",
      "Epoch 30/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3004 - acc: 0.6807 - val_loss: 0.3398 - val_acc: 0.6779\n",
      "Epoch 31/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3000 - acc: 0.6805 - val_loss: 0.3353 - val_acc: 0.6846\n",
      "Epoch 32/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2990 - acc: 0.6815 - val_loss: 0.3309 - val_acc: 0.6805\n",
      "Epoch 33/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2999 - acc: 0.6811 - val_loss: 0.3485 - val_acc: 0.6785\n",
      "Epoch 34/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2998 - acc: 0.6811 - val_loss: 0.3301 - val_acc: 0.6751\n",
      "Epoch 35/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2982 - acc: 0.6823 - val_loss: 0.3308 - val_acc: 0.6773\n",
      "Epoch 36/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2991 - acc: 0.6812 - val_loss: 0.3322 - val_acc: 0.6819\n",
      "Epoch 37/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2984 - acc: 0.6818 - val_loss: 0.3313 - val_acc: 0.6806\n",
      "Epoch 38/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2995 - acc: 0.6816 - val_loss: 0.3443 - val_acc: 0.6692\n",
      "Epoch 39/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2972 - acc: 0.6820 - val_loss: 0.3303 - val_acc: 0.6802\n",
      "Epoch 40/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2985 - acc: 0.6827 - val_loss: 0.3294 - val_acc: 0.6843\n",
      "Epoch 41/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2984 - acc: 0.6824 - val_loss: 0.3312 - val_acc: 0.6741\n",
      "Epoch 42/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2990 - acc: 0.6821 - val_loss: 0.3335 - val_acc: 0.6885\n",
      "Epoch 43/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2998 - acc: 0.6821 - val_loss: 0.3304 - val_acc: 0.6893\n",
      "Epoch 44/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2976 - acc: 0.6828 - val_loss: 0.3341 - val_acc: 0.6876\n",
      "Epoch 45/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2983 - acc: 0.6830 - val_loss: 0.3280 - val_acc: 0.6866\n",
      "Epoch 46/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2982 - acc: 0.6836 - val_loss: 0.3285 - val_acc: 0.6857\n",
      "Epoch 47/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2982 - acc: 0.6823 - val_loss: 0.3337 - val_acc: 0.6776\n",
      "Epoch 48/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2982 - acc: 0.6830 - val_loss: 0.3362 - val_acc: 0.6610\n",
      "Epoch 49/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2977 - acc: 0.6833 - val_loss: 0.3282 - val_acc: 0.6882\n",
      "Epoch 50/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3003 - acc: 0.6815 - val_loss: 0.3458 - val_acc: 0.6709\n",
      "Epoch 51/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2989 - acc: 0.6832 - val_loss: 0.3462 - val_acc: 0.6852\n",
      "Epoch 52/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2991 - acc: 0.6824 - val_loss: 0.3306 - val_acc: 0.6897\n",
      "Epoch 53/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2984 - acc: 0.6830 - val_loss: 0.3325 - val_acc: 0.6823\n",
      "Epoch 54/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2989 - acc: 0.6824 - val_loss: 0.3361 - val_acc: 0.6807\n",
      "Epoch 55/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2983 - acc: 0.6839 - val_loss: 0.3274 - val_acc: 0.6795\n",
      "Epoch 56/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3013 - acc: 0.6818 - val_loss: 0.3311 - val_acc: 0.6862\n",
      "Epoch 57/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2988 - acc: 0.6846 - val_loss: 0.3326 - val_acc: 0.6889\n",
      "Epoch 58/500\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3013 - acc: 0.6822 - val_loss: 0.3409 - val_acc: 0.6805\n",
      "Epoch 59/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2985 - acc: 0.6836 - val_loss: 0.3308 - val_acc: 0.6843\n",
      "Epoch 60/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2984 - acc: 0.6832 - val_loss: 0.3372 - val_acc: 0.6812\n",
      "Epoch 61/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2992 - acc: 0.6821 - val_loss: 0.3431 - val_acc: 0.6841\n",
      "Epoch 62/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2987 - acc: 0.6844 - val_loss: 0.3410 - val_acc: 0.6818\n",
      "Epoch 63/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2997 - acc: 0.6827 - val_loss: 0.3394 - val_acc: 0.6697\n",
      "Epoch 64/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2983 - acc: 0.6835 - val_loss: 0.3327 - val_acc: 0.6840\n",
      "Epoch 65/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2985 - acc: 0.6840 - val_loss: 0.3301 - val_acc: 0.6887\n",
      "Epoch 66/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2976 - acc: 0.6844 - val_loss: 0.3316 - val_acc: 0.6838\n",
      "Epoch 67/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2978 - acc: 0.6839 - val_loss: 0.3291 - val_acc: 0.6912\n",
      "Epoch 68/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2974 - acc: 0.6854 - val_loss: 0.3324 - val_acc: 0.6862\n",
      "Epoch 69/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2984 - acc: 0.6850 - val_loss: 0.3364 - val_acc: 0.6748\n",
      "Epoch 70/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2995 - acc: 0.6834 - val_loss: 0.3390 - val_acc: 0.6825\n",
      "Epoch 71/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2985 - acc: 0.6840 - val_loss: 0.3333 - val_acc: 0.6830\n",
      "Epoch 72/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2995 - acc: 0.6837 - val_loss: 0.3364 - val_acc: 0.6865\n",
      "Epoch 73/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2993 - acc: 0.6834 - val_loss: 0.3344 - val_acc: 0.6838\n",
      "Epoch 74/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2982 - acc: 0.6843 - val_loss: 0.3517 - val_acc: 0.6108\n",
      "Epoch 75/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2979 - acc: 0.6843 - val_loss: 0.3305 - val_acc: 0.6815\n",
      "Epoch 76/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2988 - acc: 0.6843 - val_loss: 0.3348 - val_acc: 0.6819\n",
      "Epoch 77/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2999 - acc: 0.6853 - val_loss: 0.3327 - val_acc: 0.6838\n",
      "Epoch 78/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2983 - acc: 0.6851 - val_loss: 0.3364 - val_acc: 0.6897\n",
      "Epoch 79/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2989 - acc: 0.6851 - val_loss: 0.3289 - val_acc: 0.6904\n",
      "Epoch 80/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.3002 - acc: 0.6833 - val_loss: 0.3358 - val_acc: 0.6790\n",
      "Epoch 81/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2983 - acc: 0.6843 - val_loss: 0.3474 - val_acc: 0.6642\n",
      "Epoch 82/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3001 - acc: 0.6836 - val_loss: 0.3294 - val_acc: 0.6873\n",
      "Epoch 83/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2995 - acc: 0.6836 - val_loss: 0.3346 - val_acc: 0.6887\n",
      "Epoch 84/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2976 - acc: 0.6853 - val_loss: 0.3346 - val_acc: 0.6847\n",
      "Epoch 85/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2992 - acc: 0.6847 - val_loss: 0.3348 - val_acc: 0.6789\n",
      "Epoch 86/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2985 - acc: 0.6855 - val_loss: 0.3493 - val_acc: 0.6661\n",
      "Epoch 87/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.3007 - acc: 0.6845 - val_loss: 0.3304 - val_acc: 0.6898\n",
      "Epoch 88/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2988 - acc: 0.6842 - val_loss: 0.3388 - val_acc: 0.6740\n",
      "Epoch 89/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2971 - acc: 0.6848 - val_loss: 0.3255 - val_acc: 0.6874\n",
      "Epoch 90/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2967 - acc: 0.6848 - val_loss: 0.3358 - val_acc: 0.6806\n",
      "Epoch 91/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2976 - acc: 0.6849 - val_loss: 0.3410 - val_acc: 0.6815\n",
      "Epoch 92/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2986 - acc: 0.6840 - val_loss: 0.3286 - val_acc: 0.6846\n",
      "Epoch 93/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2999 - acc: 0.6842 - val_loss: 0.3279 - val_acc: 0.6914\n",
      "Epoch 94/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2972 - acc: 0.6854 - val_loss: 0.3279 - val_acc: 0.6906\n",
      "Epoch 95/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2970 - acc: 0.6849 - val_loss: 0.3286 - val_acc: 0.6906\n",
      "Epoch 96/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2987 - acc: 0.6831 - val_loss: 0.3273 - val_acc: 0.6919\n",
      "Epoch 97/500\n",
      "227451/227451 [==============================] - 21s 93us/step - loss: 0.2959 - acc: 0.6857 - val_loss: 0.3270 - val_acc: 0.6887\n",
      "Epoch 98/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2978 - acc: 0.6848 - val_loss: 0.3306 - val_acc: 0.6890\n",
      "Epoch 99/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2978 - acc: 0.6852 - val_loss: 0.3268 - val_acc: 0.6922\n",
      "Epoch 100/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2971 - acc: 0.6860 - val_loss: 0.3270 - val_acc: 0.6936\n",
      "Epoch 101/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2963 - acc: 0.6858 - val_loss: 0.3296 - val_acc: 0.6882\n",
      "Epoch 102/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2979 - acc: 0.6843 - val_loss: 0.3344 - val_acc: 0.6833\n",
      "Epoch 103/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2965 - acc: 0.6848 - val_loss: 0.3340 - val_acc: 0.6786\n",
      "Epoch 104/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2979 - acc: 0.6853 - val_loss: 0.3375 - val_acc: 0.6772\n",
      "Epoch 105/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2977 - acc: 0.6856 - val_loss: 0.3497 - val_acc: 0.6451\n",
      "Epoch 106/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2966 - acc: 0.6853 - val_loss: 0.3310 - val_acc: 0.6807\n",
      "Epoch 107/500\n",
      "227451/227451 [==============================] - 21s 93us/step - loss: 0.2967 - acc: 0.6852 - val_loss: 0.3328 - val_acc: 0.6764\n",
      "Epoch 108/500\n",
      "227451/227451 [==============================] - 21s 94us/step - loss: 0.2959 - acc: 0.6861 - val_loss: 0.3263 - val_acc: 0.6888\n",
      "Epoch 109/500\n",
      "227451/227451 [==============================] - 21s 93us/step - loss: 0.2957 - acc: 0.6860 - val_loss: 0.3289 - val_acc: 0.6890\n",
      "Epoch 110/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2972 - acc: 0.6856 - val_loss: 0.3291 - val_acc: 0.6910\n",
      "Epoch 111/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2975 - acc: 0.6849 - val_loss: 0.3300 - val_acc: 0.6786\n",
      "Epoch 112/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2971 - acc: 0.6863 - val_loss: 0.3264 - val_acc: 0.6925\n",
      "Epoch 113/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2962 - acc: 0.6861 - val_loss: 0.3298 - val_acc: 0.6843\n",
      "Epoch 114/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2980 - acc: 0.6837 - val_loss: 0.3279 - val_acc: 0.6963\n",
      "Epoch 115/500\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2970 - acc: 0.6860 - val_loss: 0.3269 - val_acc: 0.6909\n",
      "Epoch 116/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2964 - acc: 0.6853 - val_loss: 0.3397 - val_acc: 0.6891\n",
      "Epoch 117/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2958 - acc: 0.6868 - val_loss: 0.3432 - val_acc: 0.6793\n",
      "Epoch 118/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2980 - acc: 0.6851 - val_loss: 0.3282 - val_acc: 0.6892\n",
      "Epoch 119/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2953 - acc: 0.6871 - val_loss: 0.3290 - val_acc: 0.6854\n",
      "Epoch 120/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2963 - acc: 0.6862 - val_loss: 0.3342 - val_acc: 0.6837\n",
      "Epoch 121/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2960 - acc: 0.6867 - val_loss: 0.3250 - val_acc: 0.6887\n",
      "Epoch 122/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2975 - acc: 0.6862 - val_loss: 0.3294 - val_acc: 0.6905\n",
      "Epoch 123/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2949 - acc: 0.6870 - val_loss: 0.3341 - val_acc: 0.6824\n",
      "Epoch 124/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2969 - acc: 0.6868 - val_loss: 0.3303 - val_acc: 0.6896\n",
      "Epoch 125/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2951 - acc: 0.6875 - val_loss: 0.3323 - val_acc: 0.6939\n",
      "Epoch 126/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2983 - acc: 0.6847 - val_loss: 0.3305 - val_acc: 0.6933\n",
      "Epoch 127/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2950 - acc: 0.6870 - val_loss: 0.3299 - val_acc: 0.6869\n",
      "Epoch 128/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2954 - acc: 0.6869 - val_loss: 0.3311 - val_acc: 0.6834\n",
      "Epoch 129/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2955 - acc: 0.6873 - val_loss: 0.3262 - val_acc: 0.6928\n",
      "Epoch 130/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2979 - acc: 0.6855 - val_loss: 0.3242 - val_acc: 0.6959\n",
      "Epoch 131/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2947 - acc: 0.6864 - val_loss: 0.3320 - val_acc: 0.6796\n",
      "Epoch 132/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2947 - acc: 0.6865 - val_loss: 0.3268 - val_acc: 0.6934\n",
      "Epoch 133/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2955 - acc: 0.6873 - val_loss: 0.3284 - val_acc: 0.6864\n",
      "Epoch 134/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2959 - acc: 0.6862 - val_loss: 0.3278 - val_acc: 0.6924\n",
      "Epoch 135/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2965 - acc: 0.6863 - val_loss: 0.3285 - val_acc: 0.6924\n",
      "Epoch 136/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2953 - acc: 0.6866 - val_loss: 0.3264 - val_acc: 0.6942\n",
      "Epoch 137/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2956 - acc: 0.6864 - val_loss: 0.3289 - val_acc: 0.6808\n",
      "Epoch 138/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2962 - acc: 0.6854 - val_loss: 0.3282 - val_acc: 0.6934\n",
      "Epoch 139/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2952 - acc: 0.6865 - val_loss: 0.3299 - val_acc: 0.6925\n",
      "Epoch 140/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2956 - acc: 0.6865 - val_loss: 0.3307 - val_acc: 0.6851\n",
      "Epoch 141/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2957 - acc: 0.6855 - val_loss: 0.3388 - val_acc: 0.6592\n",
      "Epoch 142/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2948 - acc: 0.6858 - val_loss: 0.3285 - val_acc: 0.6844\n",
      "Epoch 143/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2945 - acc: 0.6872 - val_loss: 0.3279 - val_acc: 0.6805\n",
      "Epoch 144/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2950 - acc: 0.6864 - val_loss: 0.3254 - val_acc: 0.6919\n",
      "Epoch 145/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2958 - acc: 0.6861 - val_loss: 0.3261 - val_acc: 0.6719\n",
      "Epoch 146/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2961 - acc: 0.6864 - val_loss: 0.3445 - val_acc: 0.6747\n",
      "Epoch 147/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2975 - acc: 0.6856 - val_loss: 0.3267 - val_acc: 0.6859\n",
      "Epoch 148/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2949 - acc: 0.6866 - val_loss: 0.3279 - val_acc: 0.6891\n",
      "Epoch 149/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2943 - acc: 0.6864 - val_loss: 0.3401 - val_acc: 0.6791\n",
      "Epoch 150/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2943 - acc: 0.6872 - val_loss: 0.3431 - val_acc: 0.6863\n",
      "Epoch 151/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2952 - acc: 0.6859 - val_loss: 0.3309 - val_acc: 0.6832\n",
      "Epoch 152/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2941 - acc: 0.6868 - val_loss: 0.3257 - val_acc: 0.6892\n",
      "Epoch 153/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2965 - acc: 0.6858 - val_loss: 0.3279 - val_acc: 0.6845\n",
      "Epoch 154/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2938 - acc: 0.6872 - val_loss: 0.3270 - val_acc: 0.6909\n",
      "Epoch 155/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2945 - acc: 0.6864 - val_loss: 0.3288 - val_acc: 0.6752\n",
      "Epoch 156/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2942 - acc: 0.6860 - val_loss: 0.3264 - val_acc: 0.6825\n",
      "Epoch 157/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2950 - acc: 0.6858 - val_loss: 0.3248 - val_acc: 0.6940\n",
      "Epoch 158/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2950 - acc: 0.6866 - val_loss: 0.3276 - val_acc: 0.6851\n",
      "Epoch 159/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2946 - acc: 0.6872 - val_loss: 0.3304 - val_acc: 0.6866\n",
      "Epoch 160/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2942 - acc: 0.6862 - val_loss: 0.3328 - val_acc: 0.6873\n",
      "Epoch 161/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2942 - acc: 0.6874 - val_loss: 0.3336 - val_acc: 0.6903\n",
      "Epoch 162/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2950 - acc: 0.6852 - val_loss: 0.3263 - val_acc: 0.6854\n",
      "Epoch 163/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2947 - acc: 0.6863 - val_loss: 0.3241 - val_acc: 0.6848\n",
      "Epoch 164/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2935 - acc: 0.6870 - val_loss: 0.3270 - val_acc: 0.6829\n",
      "Epoch 165/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2937 - acc: 0.6865 - val_loss: 0.3327 - val_acc: 0.6773\n",
      "Epoch 166/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2951 - acc: 0.6862 - val_loss: 0.3261 - val_acc: 0.6777\n",
      "Epoch 167/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2937 - acc: 0.6866 - val_loss: 0.3278 - val_acc: 0.6843\n",
      "Epoch 168/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2960 - acc: 0.6848 - val_loss: 0.3366 - val_acc: 0.6813\n",
      "Epoch 169/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2945 - acc: 0.6862 - val_loss: 0.3251 - val_acc: 0.6934\n",
      "Epoch 170/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2940 - acc: 0.6864 - val_loss: 0.3420 - val_acc: 0.6773\n",
      "Epoch 171/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2963 - acc: 0.6860 - val_loss: 0.3278 - val_acc: 0.6941\n",
      "Epoch 172/500\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2954 - acc: 0.6860 - val_loss: 0.3283 - val_acc: 0.6842\n",
      "Epoch 173/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2953 - acc: 0.6855 - val_loss: 0.3288 - val_acc: 0.6806\n",
      "Epoch 174/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2937 - acc: 0.6871 - val_loss: 0.3240 - val_acc: 0.6918\n",
      "Epoch 175/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2943 - acc: 0.6863 - val_loss: 0.3231 - val_acc: 0.6904\n",
      "Epoch 176/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2948 - acc: 0.6862 - val_loss: 0.3378 - val_acc: 0.6817\n",
      "Epoch 177/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2940 - acc: 0.6864 - val_loss: 0.3258 - val_acc: 0.6819\n",
      "Epoch 178/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2947 - acc: 0.6863 - val_loss: 0.3242 - val_acc: 0.6915\n",
      "Epoch 179/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2950 - acc: 0.6855 - val_loss: 0.3283 - val_acc: 0.6897\n",
      "Epoch 180/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2949 - acc: 0.6868 - val_loss: 0.3285 - val_acc: 0.6848\n",
      "Epoch 181/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2935 - acc: 0.6872 - val_loss: 0.3306 - val_acc: 0.6794\n",
      "Epoch 182/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2936 - acc: 0.6857 - val_loss: 0.3274 - val_acc: 0.6941\n",
      "Epoch 183/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2949 - acc: 0.6865 - val_loss: 0.3338 - val_acc: 0.6784\n",
      "Epoch 184/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2945 - acc: 0.6868 - val_loss: 0.3238 - val_acc: 0.6888\n",
      "Epoch 185/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2958 - acc: 0.6854 - val_loss: 0.3368 - val_acc: 0.6885\n",
      "Epoch 186/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2930 - acc: 0.6872 - val_loss: 0.3329 - val_acc: 0.6826\n",
      "Epoch 187/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2945 - acc: 0.6867 - val_loss: 0.3238 - val_acc: 0.6873\n",
      "Epoch 188/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2941 - acc: 0.6868 - val_loss: 0.3688 - val_acc: 0.6439\n",
      "Epoch 189/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2946 - acc: 0.6865 - val_loss: 0.3254 - val_acc: 0.6916\n",
      "Epoch 190/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2954 - acc: 0.6855 - val_loss: 0.3252 - val_acc: 0.6837\n",
      "Epoch 191/500\n",
      "227451/227451 [==============================] - 20s 88us/step - loss: 0.2942 - acc: 0.6866 - val_loss: 0.3312 - val_acc: 0.6819\n",
      "Epoch 192/500\n",
      "227451/227451 [==============================] - 20s 89us/step - loss: 0.2940 - acc: 0.6866 - val_loss: 0.3517 - val_acc: 0.6741\n",
      "Epoch 193/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2942 - acc: 0.6856 - val_loss: 0.3262 - val_acc: 0.6933\n",
      "Epoch 194/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2940 - acc: 0.6869 - val_loss: 0.3251 - val_acc: 0.6843\n",
      "Epoch 195/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2935 - acc: 0.6869 - val_loss: 0.3253 - val_acc: 0.6905\n",
      "Epoch 196/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2937 - acc: 0.6864 - val_loss: 0.3285 - val_acc: 0.6835\n",
      "Epoch 197/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2946 - acc: 0.6864 - val_loss: 0.3273 - val_acc: 0.6881\n",
      "Epoch 198/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2933 - acc: 0.6866 - val_loss: 0.3265 - val_acc: 0.6839\n",
      "Epoch 199/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2931 - acc: 0.6851 - val_loss: 0.3296 - val_acc: 0.6842\n",
      "Epoch 200/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2936 - acc: 0.6868 - val_loss: 0.3275 - val_acc: 0.6836\n",
      "Epoch 201/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2946 - acc: 0.6856 - val_loss: 0.3353 - val_acc: 0.6799\n",
      "Epoch 202/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2947 - acc: 0.6868 - val_loss: 0.3253 - val_acc: 0.6829\n",
      "Epoch 203/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2939 - acc: 0.6862 - val_loss: 0.3299 - val_acc: 0.6654\n",
      "Epoch 204/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2937 - acc: 0.6874 - val_loss: 0.3272 - val_acc: 0.6924\n",
      "Epoch 205/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2960 - acc: 0.6862 - val_loss: 0.3322 - val_acc: 0.6912\n",
      "Epoch 206/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2939 - acc: 0.6868 - val_loss: 0.3267 - val_acc: 0.6872\n",
      "Epoch 207/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2952 - acc: 0.6850 - val_loss: 0.3272 - val_acc: 0.6796\n",
      "Epoch 208/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2935 - acc: 0.6862 - val_loss: 0.3303 - val_acc: 0.6716\n",
      "Epoch 209/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2935 - acc: 0.6859 - val_loss: 0.3287 - val_acc: 0.6886\n",
      "Epoch 210/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2940 - acc: 0.6856 - val_loss: 0.3314 - val_acc: 0.6816\n",
      "Epoch 211/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2936 - acc: 0.6866 - val_loss: 0.3257 - val_acc: 0.6908\n",
      "Epoch 212/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2936 - acc: 0.6869 - val_loss: 0.3254 - val_acc: 0.6854\n",
      "Epoch 213/500\n",
      "227451/227451 [==============================] - 20s 90us/step - loss: 0.2945 - acc: 0.6865 - val_loss: 0.3261 - val_acc: 0.6913\n",
      "Epoch 214/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2944 - acc: 0.6874 - val_loss: 0.3221 - val_acc: 0.6900\n",
      "Epoch 215/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2941 - acc: 0.6863 - val_loss: 0.3315 - val_acc: 0.6748\n",
      "Epoch 216/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2936 - acc: 0.6860 - val_loss: 0.3254 - val_acc: 0.6933\n",
      "Epoch 217/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2941 - acc: 0.6852 - val_loss: 0.3424 - val_acc: 0.6713\n",
      "Epoch 218/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2940 - acc: 0.6855 - val_loss: 0.3241 - val_acc: 0.6957\n",
      "Epoch 219/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2948 - acc: 0.6872 - val_loss: 0.3273 - val_acc: 0.6895\n",
      "Epoch 220/500\n",
      "227451/227451 [==============================] - 21s 90us/step - loss: 0.2955 - acc: 0.6853 - val_loss: 0.3254 - val_acc: 0.6904\n",
      "Epoch 221/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2936 - acc: 0.6874 - val_loss: 0.3237 - val_acc: 0.6824\n",
      "Epoch 222/500\n",
      "227451/227451 [==============================] - 21s 91us/step - loss: 0.2946 - acc: 0.6864 - val_loss: 0.3340 - val_acc: 0.6844\n",
      "Epoch 223/500\n",
      "227451/227451 [==============================] - 21s 92us/step - loss: 0.2946 - acc: 0.6868 - val_loss: 0.3263 - val_acc: 0.6809\n",
      "Epoch 224/500\n",
      "106560/227451 [=============>................] - ETA: 10s - loss: 0.2883 - acc: 0.6868"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-89-41afcab9d601>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     11\u001b[0m                  \u001b[0mshuffle\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     12\u001b[0m                  \u001b[0mverbose\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 13\u001b[1;33m                  validation_data=(X_test,X_test)).history\n\u001b[0m",
      "\u001b[1;32m~\\AppData\\Local\\conda\\conda\\envs\\mlenv\\lib\\site-packages\\keras\\engine\\training.py\u001b[0m in \u001b[0;36mfit\u001b[1;34m(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, **kwargs)\u001b[0m\n\u001b[0;32m   1035\u001b[0m                                         \u001b[0minitial_epoch\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0minitial_epoch\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1036\u001b[0m                                         \u001b[0msteps_per_epoch\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0msteps_per_epoch\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1037\u001b[1;33m                                         validation_steps=validation_steps)\n\u001b[0m\u001b[0;32m   1038\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1039\u001b[0m     def evaluate(self, x=None, y=None,\n",
      "\u001b[1;32m~\\AppData\\Local\\conda\\conda\\envs\\mlenv\\lib\\site-packages\\keras\\engine\\training_arrays.py\u001b[0m in \u001b[0;36mfit_loop\u001b[1;34m(model, f, ins, out_labels, batch_size, epochs, verbose, callbacks, val_f, val_ins, shuffle, callback_metrics, initial_epoch, steps_per_epoch, validation_steps)\u001b[0m\n\u001b[0;32m    197\u001b[0m                     \u001b[0mins_batch\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mi\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mins_batch\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mi\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtoarray\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    198\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 199\u001b[1;33m                 \u001b[0mouts\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mins_batch\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    200\u001b[0m                 \u001b[0mouts\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mto_list\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mouts\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    201\u001b[0m                 \u001b[1;32mfor\u001b[0m \u001b[0ml\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mo\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mout_labels\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mouts\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\AppData\\Local\\conda\\conda\\envs\\mlenv\\lib\\site-packages\\keras\\backend\\tensorflow_backend.py\u001b[0m in \u001b[0;36m__call__\u001b[1;34m(self, inputs)\u001b[0m\n\u001b[0;32m   2664\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_legacy_call\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2665\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2666\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_call\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2667\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2668\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mpy_any\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mis_tensor\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mx\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0mx\u001b[0m \u001b[1;32min\u001b[0m \u001b[0minputs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\AppData\\Local\\conda\\conda\\envs\\mlenv\\lib\\site-packages\\keras\\backend\\tensorflow_backend.py\u001b[0m in \u001b[0;36m_call\u001b[1;34m(self, inputs)\u001b[0m\n\u001b[0;32m   2634\u001b[0m                                 \u001b[0msymbol_vals\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2635\u001b[0m                                 session)\n\u001b[1;32m-> 2636\u001b[1;33m         \u001b[0mfetched\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_callable_fn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0marray_vals\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2637\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mfetched\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0moutputs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2638\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\AppData\\Local\\conda\\conda\\envs\\mlenv\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36m__call__\u001b[1;34m(self, *args, **kwargs)\u001b[0m\n\u001b[0;32m   1380\u001b[0m           ret = tf_session.TF_SessionRunCallable(\n\u001b[0;32m   1381\u001b[0m               \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_session\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_session\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_handle\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0margs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mstatus\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1382\u001b[1;33m               run_metadata_ptr)\n\u001b[0m\u001b[0;32m   1383\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1384\u001b[0m           \u001b[0mproto_data\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "auto_encoder.compile(optimizer='adam',loss='mean_squared_error',metrics= ['accuracy'])\n",
    "\n",
    "modelcheckpoint = ModelCheckpoint(filepath='save.h5',save_best_only=True)\n",
    "\n",
    "tensorboard = TensorBoard(log_dir='./log',histogram_freq=0)\n",
    "\n",
    "History = auto_encoder.fit(X_train,X_train, \n",
    "                 batch_size=32, \n",
    "                 epochs=500, \n",
    "                 callbacks=[tensorboard, modelcheckpoint],\n",
    "                 shuffle=True,\n",
    "                 verbose=1,\n",
    "                 validation_data=(X_test,X_test)).history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = load_model('save.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x24bac888390>"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(History['loss'])\n",
    "plt.plot(History['val_loss'])\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.title('Erorr plot')\n",
    "plt.legend(['train','val'], loc='upper right')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prediction\n",
    "Prediction - Since our model has learned to give very low error on normal transaction. When we pass a fraud transaction \n",
    "ideally it should give higher error. So, we set a threshold value if any transaction is higher that this value, \n",
    "we flag it as fraud."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = model.predict(X_test)\n",
    "#error\n",
    "error = np.mean(np.power(X_test-predictions,2),axis=1)\n",
    "error_df = pd.DataFrame({\n",
    "    'error' : error,\n",
    "    'ground_truth' : y_test\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Precision, recall curve\n",
    "precision, recall, th = precision_recall_curve(error_df.ground_truth, error_df.error)\n",
    "plt.plot(recall, precision, 'b', label='Precision-Recall curve')\n",
    "plt.title('Recall vs Precision')\n",
    "plt.xlabel('Recall')\n",
    "plt.ylabel('Precision')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "threshold = 2.5\n",
    "y_pred = [1 if e > threshold else 0 for e in error_df.error.values]\n",
    "\n",
    "cmatrix = confusion_matrix(error_df.ground_truth.values , np.array(y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.heatmap(cmatrix, xticklabels=LABELS, yticklabels=LABELS, annot=True, fmt=\"d\");\n",
    "plt.title(\"Confusion matrix\")\n",
    "plt.ylabel('True class')\n",
    "plt.xlabel('Predicted class')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "'''The model was able to catch a majority of fraud detection cases (78/98) in test data. But it also flaged a lot of normal cases(377) as fraud. '''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}