DeepLearning-IBOVESPA.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "# Data Preprocessing\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import keras\n",
    "from keras import layers\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from scipy.stats.mstats import winsorize\n",
    "from sklearn.compose import ColumnTransformer\n",
    "from sklearn.preprocessing import LabelEncoder, OneHotEncoder\n",
    "from sklearn.compose import make_column_transformer\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Read the input file\n",
    "dataset = pd.read_csv('BovespaDeepLearning.csv', sep=';')\n",
    "dataset = dataset[['TICKER', 'Data', 'DATA_REGISTRO',\n",
    "       'SETOR', 'CRESC_LL_12M','CRESC_RL_12M', 'DY_12M', 'ROIC', 'ROA', 'ROE', 'Margem_Operacional',\n",
    "       'Margem_Liquida', 'Margem_Bruta', 'PSR', 'PVPA', 'IPL', 'PEG', 'LPA', 'FREEFLOAT', 'PL', 'CE', 'LG',\n",
    "       'CCL', 'LC', 'LI', 'LS', 'G', 'ICP', 'AT', 'MARKET_VALUE', 'AC', 'PC',\n",
    "       'MARKET_VALUE_AT', 'RLP_AT', 'ELP_AT', 'IMOBILIZADO_AT',\n",
    "       'CONTAS_A_RECEBER_CP_AT', 'LOP_GER_AT', 'LUCRO_PREJ_ACUM_AT', 'DIVIDA_BRUTA_AT',\n",
    "       'DIVIDA_LIQUIDA_AT', 'DESPESAS_RECEITAS_OPERACIONAIS_AT', 'CCL_AT', 'Rent_1a']]\n",
    "\n",
    "#Remove lines with nulls for the main indexes\n",
    "dataset = dataset[dataset['IPL'].notnull()]\n",
    "dataset = dataset[dataset['PVPA'].notnull()]\n",
    "dataset = dataset[dataset['AT'].notnull()]\n",
    "dataset = dataset[dataset['SETOR'].notnull()]\n",
    "dataset = dataset[dataset['LUCRO_PREJ_ACUM_AT'].notnull()]\n",
    "dataset['Ticker_ID'] = LabelEncoder().fit_transform(dataset['TICKER'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "#For the case the mean of the ticker is equal to zero\n",
    "#We will fill the null values equals to the mean of the column\n",
    "dataset = dataset.fillna(dataset.mean())\n",
    "#dataset.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "numerical=dataset.select_dtypes(include=[np.number]).columns.tolist()\n",
    "#dataset[numerical].hist(bins=20, figsize=(20, 20))\n",
    "#plt.title('Histograms')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "### CLEANING DATA\n",
    "\n",
    "# Winsorization approach to handle extreme values in the limit\n",
    "a = 0\n",
    "for i in numerical:\n",
    "    #Remove from the winsorization the y variables and histograms that seems ok\n",
    "    if numerical[a] not in ['Rent_1a','FREEFLOAT','CE'] : \n",
    "        Win = winsorize( dataset[numerical[a]],limits = [0.01,0.01])\n",
    "        dataset[numerical[a]] = Win\n",
    "    a = a + 1\n",
    "    \n",
    "dataset['Ano'] = dataset['Data'].astype(str).str[:4]\n",
    "dataset['Mes'] = dataset['Data'].astype(str).str[6:7]\n",
    "dataset = dataset[dataset['Mes'] == '6']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "898"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = dataset[dataset['Ano'] !='2019']['Rent_1a']\n",
    "a[a <= 4.45].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "898"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a[a > 4.45].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "#dataset[numerical].hist(bins=20, figsize=(20, 20))\n",
    "#plt.title('Histograms')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset['Rentability_1a'] = dataset['Rent_1a']\n",
    "\n",
    "# Set target rentability %\n",
    "buy = 4.45\n",
    "\n",
    "def f(row):\n",
    "    if row['Rent_1a'] > buy:\n",
    "        val = 1\n",
    "    else:\n",
    "        val = 0\n",
    "    return val\n",
    "dataset['Rent_1a'] = dataset.apply(f, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Here we ignore the first 3 (non-predictive features)\n",
    "X = dataset.iloc[:, 3:49]\n",
    "# y is a vector containing the information about the target\n",
    "y = dataset.iloc[:, 43:48].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "column_trans = make_column_transformer(\n",
    "    (OneHotEncoder(), ['SETOR']),\n",
    "    remainder='passthrough')\n",
    "X = column_trans.fit_transform(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Xf = X[X[:,58]=='2019']\n",
    "Xa = X[X[:,58]!='2019']\n",
    "\n",
    "#yf = y[y[:,2]=='2019']\n",
    "ya = y[y[:,2]!='2019']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creation of training and test sets\n",
    "from sklearn.model_selection import train_test_split\n",
    "X_train_c, X_test_c, y_train_c, y_test_c = train_test_split(Xa, ya, test_size = 0.5, random_state=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = y_train_c[:,0].astype(str).astype('int64')\n",
    "y_test = y_test_c[:,0].astype(str).astype('int64')\n",
    "#yf = yf[:,0].astype(str).astype(int)\n",
    "X_train = X_train_c[:,:56]\n",
    "X_test = X_test_c[:,:56]\n",
    "#Xf = Xf[:,:56]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Preprocessing: feature normalization\n",
    "#mean = X_train.mean(axis=0)\n",
    "#X_train -= mean\n",
    "#std = X_train.std(axis=0)\n",
    "#X_train /= std\n",
    "\n",
    "#X_test -= mean\n",
    "#X_test /= std\n",
    "\n",
    "\n",
    "# Preprocessing: feature Standardization\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "sc = StandardScaler()\n",
    "# Standardization only uses the information on the training set!!!\n",
    "X_train = sc.fit_transform(X_train)\n",
    "X_test = sc.transform(X_test)\n",
    "#Xf = sc.transform(Xf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Creation of the model\n",
    "\n",
    "def build_model():\n",
    "    # Creation of an empty model\n",
    "    model = Sequential()\n",
    "\n",
    "    # Adding the input layer and the first hidden layer\n",
    "    model.add(Dense(units = 5, kernel_initializer = 'uniform', activation = 'relu', input_dim = 56))\n",
    "\n",
    "    # Adding the second hidden layer\n",
    "    model.add(Dense(units = 50, kernel_initializer = 'uniform', activation = 'relu'))\n",
    "    \n",
    "    # Adding the second hidden layer\n",
    "    model.add(Dense(units = 5, kernel_initializer = 'uniform', activation = 'sigmoid'))\n",
    "    \n",
    "    # Adding the output layer\n",
    "    model.add(Dense(units = 1, kernel_initializer = 'uniform', activation = 'sigmoid'))\n",
    "\n",
    "    model.compile(optimizer = 'Adam', loss = 'binary_crossentropy', metrics = ['accuracy'])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing fold # 0\n",
      "Train on 719 samples, validate on 179 samples\n",
      "Epoch 1/15\n",
      "719/719 [==============================] - 1s 1ms/step - loss: 0.6934 - accuracy: 0.4784 - val_loss: 0.6933 - val_accuracy: 0.4804\n",
      "Epoch 2/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6931 - accuracy: 0.5090 - val_loss: 0.6933 - val_accuracy: 0.4804\n",
      "Epoch 3/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6930 - accuracy: 0.5090 - val_loss: 0.6934 - val_accuracy: 0.4804\n",
      "Epoch 4/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6928 - accuracy: 0.5104 - val_loss: 0.6934 - val_accuracy: 0.4804\n",
      "Epoch 5/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6923 - accuracy: 0.5229 - val_loss: 0.6932 - val_accuracy: 0.4916\n",
      "Epoch 6/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6914 - accuracy: 0.5522 - val_loss: 0.6926 - val_accuracy: 0.5140\n",
      "Epoch 7/15\n",
      "719/719 [==============================] - 0s 67us/step - loss: 0.6901 - accuracy: 0.5828 - val_loss: 0.6915 - val_accuracy: 0.5419\n",
      "Epoch 8/15\n",
      "719/719 [==============================] - 0s 67us/step - loss: 0.6883 - accuracy: 0.6106 - val_loss: 0.6901 - val_accuracy: 0.5531\n",
      "Epoch 9/15\n",
      "719/719 [==============================] - 0s 67us/step - loss: 0.6860 - accuracy: 0.6356 - val_loss: 0.6889 - val_accuracy: 0.5866\n",
      "Epoch 10/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6834 - accuracy: 0.6467 - val_loss: 0.6876 - val_accuracy: 0.5866\n",
      "Epoch 11/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6793 - accuracy: 0.6537 - val_loss: 0.6861 - val_accuracy: 0.5531\n",
      "Epoch 12/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6746 - accuracy: 0.6718 - val_loss: 0.6848 - val_accuracy: 0.5754\n",
      "Epoch 13/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6695 - accuracy: 0.6704 - val_loss: 0.6839 - val_accuracy: 0.5587\n",
      "Epoch 14/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6627 - accuracy: 0.6801 - val_loss: 0.6820 - val_accuracy: 0.5754\n",
      "Epoch 15/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6554 - accuracy: 0.6871 - val_loss: 0.6803 - val_accuracy: 0.5810\n",
      "Processing fold # 1\n",
      "Train on 719 samples, validate on 179 samples\n",
      "Epoch 1/15\n",
      "719/719 [==============================] - 1s 1ms/step - loss: 0.6933 - accuracy: 0.4924 - val_loss: 0.6932 - val_accuracy: 0.4916\n",
      "Epoch 2/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6932 - accuracy: 0.4979 - val_loss: 0.6931 - val_accuracy: 0.4916\n",
      "Epoch 3/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6931 - accuracy: 0.4840 - val_loss: 0.6931 - val_accuracy: 0.5140\n",
      "Epoch 4/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6929 - accuracy: 0.5202 - val_loss: 0.6929 - val_accuracy: 0.5251\n",
      "Epoch 5/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6923 - accuracy: 0.5563 - val_loss: 0.6924 - val_accuracy: 0.5531\n",
      "Epoch 6/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6908 - accuracy: 0.6120 - val_loss: 0.6914 - val_accuracy: 0.5363\n",
      "Epoch 7/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6887 - accuracy: 0.6245 - val_loss: 0.6904 - val_accuracy: 0.5922\n",
      "Epoch 8/15\n",
      "719/719 [==============================] - 0s 111us/step - loss: 0.6866 - accuracy: 0.6231 - val_loss: 0.6889 - val_accuracy: 0.5866\n",
      "Epoch 9/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6837 - accuracy: 0.6314 - val_loss: 0.6870 - val_accuracy: 0.5866\n",
      "Epoch 10/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6801 - accuracy: 0.6342 - val_loss: 0.6846 - val_accuracy: 0.5978\n",
      "Epoch 11/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6756 - accuracy: 0.6453 - val_loss: 0.6818 - val_accuracy: 0.5922\n",
      "Epoch 12/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6706 - accuracy: 0.6481 - val_loss: 0.6787 - val_accuracy: 0.5810\n",
      "Epoch 13/15\n",
      "719/719 [==============================] - 0s 67us/step - loss: 0.6647 - accuracy: 0.6551 - val_loss: 0.6749 - val_accuracy: 0.5754\n",
      "Epoch 14/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6580 - accuracy: 0.6745 - val_loss: 0.6729 - val_accuracy: 0.5866\n",
      "Epoch 15/15\n",
      "719/719 [==============================] - 0s 81us/step - loss: 0.6517 - accuracy: 0.6732 - val_loss: 0.6704 - val_accuracy: 0.5866\n",
      "Processing fold # 2\n",
      "Train on 719 samples, validate on 179 samples\n",
      "Epoch 1/15\n",
      "719/719 [==============================] - 1s 1ms/step - loss: 0.6937 - accuracy: 0.4951 - val_loss: 0.6931 - val_accuracy: 0.5028\n",
      "Epoch 2/15\n",
      "719/719 [==============================] - 0s 156us/step - loss: 0.6932 - accuracy: 0.4951 - val_loss: 0.6931 - val_accuracy: 0.5028\n",
      "Epoch 3/15\n",
      "719/719 [==============================] - 0s 111us/step - loss: 0.6931 - accuracy: 0.4951 - val_loss: 0.6930 - val_accuracy: 0.5028\n",
      "Epoch 4/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6926 - accuracy: 0.5271 - val_loss: 0.6927 - val_accuracy: 0.5363\n",
      "Epoch 5/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6915 - accuracy: 0.6147 - val_loss: 0.6923 - val_accuracy: 0.5531\n",
      "Epoch 6/15\n",
      "719/719 [==============================] - 0s 111us/step - loss: 0.6892 - accuracy: 0.6314 - val_loss: 0.6921 - val_accuracy: 0.4916\n",
      "Epoch 7/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6867 - accuracy: 0.6328 - val_loss: 0.6914 - val_accuracy: 0.5307\n",
      "Epoch 8/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6829 - accuracy: 0.6356 - val_loss: 0.6907 - val_accuracy: 0.5419\n",
      "Epoch 9/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6776 - accuracy: 0.6523 - val_loss: 0.6902 - val_accuracy: 0.5363\n",
      "Epoch 10/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6718 - accuracy: 0.6537 - val_loss: 0.6906 - val_accuracy: 0.5475\n",
      "Epoch 11/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6644 - accuracy: 0.6662 - val_loss: 0.6900 - val_accuracy: 0.5531\n",
      "Epoch 12/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6574 - accuracy: 0.6732 - val_loss: 0.6901 - val_accuracy: 0.5475\n",
      "Epoch 13/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6505 - accuracy: 0.6843 - val_loss: 0.6887 - val_accuracy: 0.5475\n",
      "Epoch 14/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6428 - accuracy: 0.6940 - val_loss: 0.6883 - val_accuracy: 0.5587\n",
      "Epoch 15/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6355 - accuracy: 0.6940 - val_loss: 0.6912 - val_accuracy: 0.5363\n",
      "Processing fold # 3\n",
      "Train on 719 samples, validate on 179 samples\n",
      "Epoch 1/15\n",
      "719/719 [==============================] - 1s 1ms/step - loss: 0.6933 - accuracy: 0.4993 - val_loss: 0.6932 - val_accuracy: 0.4860\n",
      "Epoch 2/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6933 - accuracy: 0.4645 - val_loss: 0.6931 - val_accuracy: 0.5140\n",
      "Epoch 3/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6932 - accuracy: 0.4896 - val_loss: 0.6931 - val_accuracy: 0.4860\n",
      "Epoch 4/15\n",
      "719/719 [==============================] - 0s 111us/step - loss: 0.6931 - accuracy: 0.4924 - val_loss: 0.6929 - val_accuracy: 0.6034\n",
      "Epoch 5/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6927 - accuracy: 0.5883 - val_loss: 0.6924 - val_accuracy: 0.5642\n",
      "Epoch 6/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6917 - accuracy: 0.6064 - val_loss: 0.6912 - val_accuracy: 0.5922\n",
      "Epoch 7/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6898 - accuracy: 0.5981 - val_loss: 0.6887 - val_accuracy: 0.6145\n",
      "Epoch 8/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6865 - accuracy: 0.5953 - val_loss: 0.6851 - val_accuracy: 0.6313\n",
      "Epoch 9/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6823 - accuracy: 0.6203 - val_loss: 0.6819 - val_accuracy: 0.6201\n",
      "Epoch 10/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6776 - accuracy: 0.6314 - val_loss: 0.6792 - val_accuracy: 0.5754\n",
      "Epoch 11/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6731 - accuracy: 0.6342 - val_loss: 0.6757 - val_accuracy: 0.5866\n",
      "Epoch 12/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6674 - accuracy: 0.6481 - val_loss: 0.6744 - val_accuracy: 0.5810\n",
      "Epoch 13/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6622 - accuracy: 0.6509 - val_loss: 0.6718 - val_accuracy: 0.5754\n",
      "Epoch 14/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6561 - accuracy: 0.6662 - val_loss: 0.6708 - val_accuracy: 0.5922\n",
      "Epoch 15/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6502 - accuracy: 0.6745 - val_loss: 0.6677 - val_accuracy: 0.5978\n",
      "Processing fold # 4\n",
      "Train on 719 samples, validate on 179 samples\n",
      "Epoch 1/15\n",
      "719/719 [==============================] - 1s 2ms/step - loss: 0.6935 - accuracy: 0.4993 - val_loss: 0.6926 - val_accuracy: 0.5196\n",
      "Epoch 2/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6933 - accuracy: 0.4993 - val_loss: 0.6926 - val_accuracy: 0.5196\n",
      "Epoch 3/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6933 - accuracy: 0.4993 - val_loss: 0.6928 - val_accuracy: 0.5196\n",
      "Epoch 4/15\n",
      "719/719 [==============================] - 0s 134us/step - loss: 0.6931 - accuracy: 0.4993 - val_loss: 0.6927 - val_accuracy: 0.5196\n",
      "Epoch 5/15\n",
      "719/719 [==============================] - 0s 134us/step - loss: 0.6929 - accuracy: 0.5118 - val_loss: 0.6926 - val_accuracy: 0.5363\n",
      "Epoch 6/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6922 - accuracy: 0.5452 - val_loss: 0.6922 - val_accuracy: 0.5642\n",
      "Epoch 7/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6910 - accuracy: 0.5911 - val_loss: 0.6917 - val_accuracy: 0.5698\n",
      "Epoch 8/15\n",
      "719/719 [==============================] - 0s 111us/step - loss: 0.6893 - accuracy: 0.6036 - val_loss: 0.6906 - val_accuracy: 0.5698\n",
      "Epoch 9/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6865 - accuracy: 0.6259 - val_loss: 0.6894 - val_accuracy: 0.5587\n",
      "Epoch 10/15\n",
      "719/719 [==============================] - 0s 122us/step - loss: 0.6826 - accuracy: 0.6384 - val_loss: 0.6867 - val_accuracy: 0.5698\n",
      "Epoch 11/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6776 - accuracy: 0.6398 - val_loss: 0.6845 - val_accuracy: 0.5642\n",
      "Epoch 12/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6719 - accuracy: 0.6481 - val_loss: 0.6815 - val_accuracy: 0.5922\n",
      "Epoch 13/15\n",
      "719/719 [==============================] - 0s 100us/step - loss: 0.6655 - accuracy: 0.6551 - val_loss: 0.6800 - val_accuracy: 0.5866\n",
      "Epoch 14/15\n",
      "719/719 [==============================] - 0s 78us/step - loss: 0.6595 - accuracy: 0.6732 - val_loss: 0.6790 - val_accuracy: 0.5978\n",
      "Epoch 15/15\n",
      "719/719 [==============================] - 0s 89us/step - loss: 0.6536 - accuracy: 0.6759 - val_loss: 0.6778 - val_accuracy: 0.6089\n",
      "all_scores : [0.5810055732727051, 0.5865921974182129, 0.5363128781318665, 0.5977653861045837, 0.6089385747909546]\n",
      "mean all scores : 0.5821229219436646\n"
     ]
    }
   ],
   "source": [
    "# 5-fold Cross Validation \n",
    "all_acc_histories = []\n",
    "all_val_acc_histories = []\n",
    "all_loss_histories = []\n",
    "all_val_loss_histories = []\n",
    "\n",
    "k = 5\n",
    "num_val_samples = len(X_train) // k\n",
    "num_epochs = 15\n",
    "all_scores = []\n",
    "\n",
    "\n",
    "for i in range(k):\n",
    "    print(f'Processing fold # {i}')\n",
    "    val_data = X_train[i * num_val_samples: (i+1) * num_val_samples]\n",
    "    val_targets = y_train[i * num_val_samples: (i+1) * num_val_samples]\n",
    "    \n",
    "    partial_train_data = np.concatenate(\n",
    "                            [X_train[:i * num_val_samples],\n",
    "                            X_train[(i+1) * num_val_samples:]],\n",
    "                            axis=0)\n",
    "    partial_train_targets = np.concatenate(\n",
    "                            [y_train[:i * num_val_samples],\n",
    "                            y_train[(i+1)*num_val_samples:]],\n",
    "                            axis=0)\n",
    "    model = build_model()\n",
    "    \n",
    "    history = model.fit(partial_train_data,\n",
    "              partial_train_targets,\n",
    "              validation_data=(val_data, val_targets),\n",
    "              epochs=num_epochs,\n",
    "                       )\n",
    "    val_mse, val_acc = model.evaluate(val_data, val_targets, verbose=0)\n",
    "    all_scores.append(val_acc)\n",
    "    \n",
    "    acc_history = history.history['accuracy']\n",
    "    val_acc_history = history.history['val_accuracy']\n",
    "    loss_history = history.history['loss']\n",
    "    val_loss_history = history.history['val_loss']\n",
    "    \n",
    "    all_acc_histories.append(acc_history)\n",
    "    all_val_acc_histories.append(val_acc_history)\n",
    "    all_loss_histories.append(loss_history)\n",
    "    all_val_loss_histories.append(val_loss_history)\n",
    "\n",
    "\t\n",
    "# Here we print the scores\n",
    "print(f'all_scores : {all_scores}')\n",
    "print(f'mean all scores : {np.mean(all_scores)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Here we compute the history of successive mean K-fold validation scores\n",
    "average_acc_history = [np.mean([x[i] for x in all_acc_histories]) for i in range(num_epochs)]\n",
    "average_val_acc_history = [np.mean([x[i] for x in all_val_acc_histories]) for i in range(num_epochs)]\n",
    "average_loss_history = [np.mean([x[i] for x in all_loss_histories]) for i in range(num_epochs)]\n",
    "average_val_loss_history = [np.mean([x[i] for x in all_val_loss_histories]) for i in range(num_epochs)]\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "plt.plot(range(1, len(average_acc_history) + 1), average_acc_history)\n",
    "plt.plot(range(1, len(average_val_acc_history) + 1), average_val_acc_history)\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('ACC')\n",
    "plt.show()\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "plt.plot(range(1, len(average_loss_history) + 1), average_loss_history)\n",
    "plt.plot(range(1, len(average_val_loss_history) + 1), average_val_loss_history)\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "  Do Nothing       0.58      0.54      0.56       452\n",
      "         Buy       0.57      0.61      0.59       446\n",
      "\n",
      "    accuracy                           0.57       898\n",
      "   macro avg       0.57      0.57      0.57       898\n",
      "weighted avg       0.57      0.57      0.57       898\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Predicting the Test set results\n",
    "y_pred = model.predict(X_test)\n",
    "y_pred = (y_pred > 0.50).astype(int)\n",
    "\n",
    "# Making the Confusion Matrix\n",
    "from sklearn.metrics import confusion_matrix\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "\n",
    "# Print Statistics from Confusion Matrix\n",
    "from sklearn.metrics import classification_report\n",
    "target_names = ['Do Nothing','Buy']\n",
    "print(classification_report(y_test, y_pred, target_names=target_names))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: 419, 1: 479}"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique, counts = np.unique(y_pred, return_counts=True)\n",
    "dict(zip(unique, counts))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "19.87774665588597"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(y_test_c[:,4].astype(str).astype(float) * y_pred[:,0].astype(int))/(dict(zip(unique, counts))[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "13.344528607248854"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(y_test_c[:,4].astype(str).astype(float))/len(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.0 64-bit",
   "name": "python390jvsc74a57bd0727915aaf93b8fd66db277de47a9e53b4b1e8995bbde1dbbda6fae2300a5a514"
  },
  "language_info": {
   "name": "python",
   "version": ""
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}