stock_rnn_lstm.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using Theano backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from keras.layers import Dense, Dropout, Activation, LSTM, Convolution1D, MaxPooling1D\n",
    "from keras.models import Sequential\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## Data preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Date</th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Adj Close</th>\n",
       "      <th>Volume</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4271</th>\n",
       "      <td>2017-11-27</td>\n",
       "      <td>23552.859375</td>\n",
       "      <td>23638.919922</td>\n",
       "      <td>23545.019531</td>\n",
       "      <td>23580.779297</td>\n",
       "      <td>23580.779297</td>\n",
       "      <td>296320000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4272</th>\n",
       "      <td>2017-11-28</td>\n",
       "      <td>23625.189453</td>\n",
       "      <td>23849.609375</td>\n",
       "      <td>23617.109375</td>\n",
       "      <td>23836.710938</td>\n",
       "      <td>23836.710938</td>\n",
       "      <td>319240000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4273</th>\n",
       "      <td>2017-11-29</td>\n",
       "      <td>23883.259766</td>\n",
       "      <td>23959.759766</td>\n",
       "      <td>23872.589844</td>\n",
       "      <td>23940.679688</td>\n",
       "      <td>23940.679688</td>\n",
       "      <td>401410000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4274</th>\n",
       "      <td>2017-11-30</td>\n",
       "      <td>24013.800781</td>\n",
       "      <td>24327.820313</td>\n",
       "      <td>24013.800781</td>\n",
       "      <td>24272.349609</td>\n",
       "      <td>24272.349609</td>\n",
       "      <td>488200000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4275</th>\n",
       "      <td>2017-12-01</td>\n",
       "      <td>24305.400391</td>\n",
       "      <td>24322.060547</td>\n",
       "      <td>23921.900391</td>\n",
       "      <td>24231.589844</td>\n",
       "      <td>24231.589844</td>\n",
       "      <td>417910000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Date          Open          High           Low         Close  \\\n",
       "4271  2017-11-27  23552.859375  23638.919922  23545.019531  23580.779297   \n",
       "4272  2017-11-28  23625.189453  23849.609375  23617.109375  23836.710938   \n",
       "4273  2017-11-29  23883.259766  23959.759766  23872.589844  23940.679688   \n",
       "4274  2017-11-30  24013.800781  24327.820313  24013.800781  24272.349609   \n",
       "4275  2017-12-01  24305.400391  24322.060547  23921.900391  24231.589844   \n",
       "\n",
       "         Adj Close     Volume  \n",
       "4271  23580.779297  296320000  \n",
       "4272  23836.710938  319240000  \n",
       "4273  23940.679688  401410000  \n",
       "4274  24272.349609  488200000  \n",
       "4275  24231.589844  417910000  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data= pd.read_csv('/Users/arshiyaaggarwal/Desktop/Fms_Project/code/data/^DJI.csv')\n",
    "data.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Date</th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Adj Close</th>\n",
       "      <th>Volume</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4271</th>\n",
       "      <td>2017-11-27</td>\n",
       "      <td>23552.859375</td>\n",
       "      <td>23638.919922</td>\n",
       "      <td>23545.019531</td>\n",
       "      <td>23580.779297</td>\n",
       "      <td>23580.779297</td>\n",
       "      <td>296320000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4272</th>\n",
       "      <td>2017-11-28</td>\n",
       "      <td>23625.189453</td>\n",
       "      <td>23849.609375</td>\n",
       "      <td>23617.109375</td>\n",
       "      <td>23836.710938</td>\n",
       "      <td>23836.710938</td>\n",
       "      <td>319240000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4273</th>\n",
       "      <td>2017-11-29</td>\n",
       "      <td>23883.259766</td>\n",
       "      <td>23959.759766</td>\n",
       "      <td>23872.589844</td>\n",
       "      <td>23940.679688</td>\n",
       "      <td>23940.679688</td>\n",
       "      <td>401410000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4274</th>\n",
       "      <td>2017-11-30</td>\n",
       "      <td>24013.800781</td>\n",
       "      <td>24327.820313</td>\n",
       "      <td>24013.800781</td>\n",
       "      <td>24272.349609</td>\n",
       "      <td>24272.349609</td>\n",
       "      <td>488200000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4275</th>\n",
       "      <td>2017-12-01</td>\n",
       "      <td>24305.400391</td>\n",
       "      <td>24322.060547</td>\n",
       "      <td>23921.900391</td>\n",
       "      <td>24231.589844</td>\n",
       "      <td>24231.589844</td>\n",
       "      <td>417910000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Date          Open          High           Low         Close  \\\n",
       "4271  2017-11-27  23552.859375  23638.919922  23545.019531  23580.779297   \n",
       "4272  2017-11-28  23625.189453  23849.609375  23617.109375  23836.710938   \n",
       "4273  2017-11-29  23883.259766  23959.759766  23872.589844  23940.679688   \n",
       "4274  2017-11-30  24013.800781  24327.820313  24013.800781  24272.349609   \n",
       "4275  2017-12-01  24305.400391  24322.060547  23921.900391  24231.589844   \n",
       "\n",
       "         Adj Close     Volume  \n",
       "4271  23580.779297  296320000  \n",
       "4272  23836.710938  319240000  \n",
       "4273  23940.679688  401410000  \n",
       "4274  24272.349609  488200000  \n",
       "4275  24231.589844  417910000  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data= data.dropna(axis=0)\n",
    "data.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4276, 4)\n"
     ]
    }
   ],
   "source": [
    "data= data.values[:, 1:5]\n",
    "print data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[10377.330078 10604.459961 10319.30957 10560.950195]\n",
      " [10576.780273 10917.299805 10576.780273 10898.719727]\n",
      " [10896.139648 10896.139648 10620.860352 10664.379883]\n",
      " [10644.660156 10689.019531 10583.230469 10617.360352]\n",
      " [10632.139648 10803.540039 10631.860352 10712.910156]\n",
      " [10719.360352 10792.160156 10670.05957 10725.799805]]\n"
     ]
    }
   ],
   "source": [
    "print data[:6, :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4275,)\n",
      "(4275, 4)\n"
     ]
    }
   ],
   "source": [
    "y= data[1:, 3]\n",
    "print y.shape\n",
    "X= data[:4275,:]\n",
    "print X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y= (y- min(y))/(max(y)- min(y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.9609839991623761 0.9754227755909837 0.9812883322331646 1.0\n",
      " 0.9977004752838762]\n",
      "[0.2455061392540156 0.2322854972004963 0.2296328181756039\n",
      " 0.2350234070545823 0.23575059639087165 0.23814659118191125\n",
      " 0.23962306262608365]\n"
     ]
    }
   ],
   "source": [
    "print y[4270:]\n",
    "print y[:7]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4275, 4)\n",
      "[[0.2192915864574453 0.22106956373653716 0.2194135977236257\n",
      "  0.22645035256860355]\n",
      " [0.23071040945868904 0.23882618237200137 0.23408943537613414\n",
      "  0.2455061392540156]\n",
      " [0.24899421297621796 0.2376251434208518 0.23660200117712055\n",
      "  0.2322854972004963]]\n"
     ]
    }
   ],
   "source": [
    "X= (X- np.min(X,axis=0))/(np.max(X, axis=0)-np.min(X,axis=0))\n",
    "print X.shape\n",
    "print X[:3, :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "##LSTM input layer preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4176, 100, 4) (4176,)\n"
     ]
    }
   ],
   "source": [
    "time_steps= 100\n",
    "X_new= np.zeros((X.shape[0] - time_steps +1, 100, X.shape[1]))\n",
    "y_new= np.zeros((y.shape[0] -time_steps +1,))\n",
    "for ix in range(X_new.shape[0]):\n",
    "    for jx in range(time_steps):\n",
    "        X_new[ix, jx, :]= X[ix +jx, :]\n",
    "    y_new[ix]= y[ix + time_steps -1]\n",
    "print X_new.shape, y_new.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## split into train and test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3420, 100, 4) (3420,)\n",
      "(756, 100, 4) (756,)\n"
     ]
    }
   ],
   "source": [
    "split = int(0.8*data.shape[0])\n",
    "X_train = X_new[:split]\n",
    "X_test = X_new[split:]\n",
    "\n",
    "y_train = y_new[:split]\n",
    "y_test = y_new[split:]\n",
    "\n",
    "print X_train.shape, y_train.shape\n",
    "print X_test.shape, y_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## Model fitting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "____________________________________________________________________________________________________\n",
      "Layer (type)                     Output Shape          Param #     Connected to                     \n",
      "====================================================================================================\n",
      "lstm_1 (LSTM)                    (None, 100, 100)      42000       lstm_input_1[0][0]               \n",
      "____________________________________________________________________________________________________\n",
      "dropout_1 (Dropout)              (None, 100, 100)      0           lstm_1[0][0]                     \n",
      "____________________________________________________________________________________________________\n",
      "lstm_2 (LSTM)                    (None, 50)            30200       dropout_1[0][0]                  \n",
      "____________________________________________________________________________________________________\n",
      "dropout_2 (Dropout)              (None, 50)            0           lstm_2[0][0]                     \n",
      "____________________________________________________________________________________________________\n",
      "dense_1 (Dense)                  (None, 1)             51          dropout_2[0][0]                  \n",
      "____________________________________________________________________________________________________\n",
      "activation_1 (Activation)        (None, 1)             0           dense_1[0][0]                    \n",
      "====================================================================================================\n",
      "Total params: 72,251\n",
      "Trainable params: 72,251\n",
      "Non-trainable params: 0\n",
      "____________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(LSTM(100, input_shape= (100,4), return_sequences=True))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(LSTM(50, return_sequences=False))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(output_dim=1))\n",
    "model.add(Activation('linear'))\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python2.7/site-packages/theano/tensor/basic.py:2146: UserWarning: theano.tensor.round() changed its default from `half_away_from_zero` to `half_to_even` to have the same default as NumPy. Use the Theano flag `warn.round=False` to disable this warning.\n",
      "  \"theano.tensor.round() changed its default from\"\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss='mse', optimizer='rmsprop', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 3420 samples, validate on 756 samples\n",
      "Epoch 1/50\n",
      "19s - loss: 0.0558 - acc: 2.9240e-04 - val_loss: 0.0123 - val_acc: 0.0013\n",
      "Epoch 2/50\n",
      "18s - loss: 0.0052 - acc: 2.9240e-04 - val_loss: 0.0043 - val_acc: 0.0013\n",
      "Epoch 3/50\n",
      "18s - loss: 0.0055 - acc: 2.9240e-04 - val_loss: 0.0022 - val_acc: 0.0013\n",
      "Epoch 4/50\n",
      "16s - loss: 0.0044 - acc: 2.9240e-04 - val_loss: 0.0026 - val_acc: 0.0013\n",
      "Epoch 5/50\n",
      "18s - loss: 0.0044 - acc: 2.9240e-04 - val_loss: 0.0014 - val_acc: 0.0013\n",
      "Epoch 6/50\n",
      "18s - loss: 0.0031 - acc: 2.9240e-04 - val_loss: 0.0061 - val_acc: 0.0013\n",
      "Epoch 7/50\n",
      "16s - loss: 0.0027 - acc: 2.9240e-04 - val_loss: 0.0123 - val_acc: 0.0013\n",
      "Epoch 8/50\n",
      "16s - loss: 0.0022 - acc: 2.9240e-04 - val_loss: 0.0232 - val_acc: 0.0013\n",
      "Epoch 9/50\n",
      "16s - loss: 0.0022 - acc: 2.9240e-04 - val_loss: 0.0204 - val_acc: 0.0013\n",
      "Epoch 10/50\n",
      "15s - loss: 0.0019 - acc: 2.9240e-04 - val_loss: 0.0160 - val_acc: 0.0013\n",
      "Epoch 11/50\n",
      "16s - loss: 0.0020 - acc: 2.9240e-04 - val_loss: 0.0051 - val_acc: 0.0013\n",
      "Epoch 12/50\n",
      "16s - loss: 0.0019 - acc: 2.9240e-04 - val_loss: 0.0086 - val_acc: 0.0013\n",
      "Epoch 13/50\n",
      "16s - loss: 0.0016 - acc: 2.9240e-04 - val_loss: 0.0167 - val_acc: 0.0013\n",
      "Epoch 14/50\n",
      "15s - loss: 0.0018 - acc: 2.9240e-04 - val_loss: 0.0161 - val_acc: 0.0013\n",
      "Epoch 15/50\n",
      "16s - loss: 0.0014 - acc: 2.9240e-04 - val_loss: 0.0150 - val_acc: 0.0013\n",
      "Epoch 16/50\n",
      "16s - loss: 0.0015 - acc: 2.9240e-04 - val_loss: 0.0164 - val_acc: 0.0013\n",
      "Epoch 17/50\n",
      "16s - loss: 0.0013 - acc: 2.9240e-04 - val_loss: 0.0151 - val_acc: 0.0013\n",
      "Epoch 18/50\n",
      "15s - loss: 0.0013 - acc: 2.9240e-04 - val_loss: 0.0061 - val_acc: 0.0013\n",
      "Epoch 19/50\n",
      "16s - loss: 0.0013 - acc: 2.9240e-04 - val_loss: 0.0027 - val_acc: 0.0013\n",
      "Epoch 20/50\n",
      "15s - loss: 0.0013 - acc: 2.9240e-04 - val_loss: 0.0045 - val_acc: 0.0013\n",
      "Epoch 21/50\n",
      "15s - loss: 0.0013 - acc: 2.9240e-04 - val_loss: 0.0019 - val_acc: 0.0013\n",
      "Epoch 22/50\n",
      "15s - loss: 0.0012 - acc: 2.9240e-04 - val_loss: 0.0027 - val_acc: 0.0013\n",
      "Epoch 23/50\n",
      "14s - loss: 0.0013 - acc: 2.9240e-04 - val_loss: 0.0037 - val_acc: 0.0013\n",
      "Epoch 24/50\n",
      "14s - loss: 0.0011 - acc: 2.9240e-04 - val_loss: 0.0022 - val_acc: 0.0013\n",
      "Epoch 25/50\n",
      "16s - loss: 0.0011 - acc: 2.9240e-04 - val_loss: 0.0029 - val_acc: 0.0013\n",
      "Epoch 26/50\n",
      "16s - loss: 0.0011 - acc: 2.9240e-04 - val_loss: 0.0066 - val_acc: 0.0013\n",
      "Epoch 27/50\n",
      "15s - loss: 0.0010 - acc: 2.9240e-04 - val_loss: 0.0050 - val_acc: 0.0013\n",
      "Epoch 28/50\n",
      "15s - loss: 0.0010 - acc: 2.9240e-04 - val_loss: 0.0044 - val_acc: 0.0013\n",
      "Epoch 29/50\n",
      "15s - loss: 0.0011 - acc: 2.9240e-04 - val_loss: 0.0049 - val_acc: 0.0013\n",
      "Epoch 30/50\n",
      "13s - loss: 8.0355e-04 - acc: 2.9240e-04 - val_loss: 4.7968e-04 - val_acc: 0.0013\n",
      "Epoch 31/50\n",
      "13s - loss: 0.0011 - acc: 2.9240e-04 - val_loss: 9.9860e-04 - val_acc: 0.0013\n",
      "Epoch 32/50\n",
      "13s - loss: 0.0011 - acc: 2.9240e-04 - val_loss: 0.0038 - val_acc: 0.0013\n",
      "Epoch 33/50\n",
      "13s - loss: 9.2845e-04 - acc: 2.9240e-04 - val_loss: 8.9745e-04 - val_acc: 0.0013\n",
      "Epoch 34/50\n",
      "14s - loss: 8.7767e-04 - acc: 2.9240e-04 - val_loss: 5.8416e-04 - val_acc: 0.0013\n",
      "Epoch 35/50\n",
      "14s - loss: 9.2837e-04 - acc: 2.9240e-04 - val_loss: 3.7143e-04 - val_acc: 0.0013\n",
      "Epoch 36/50\n",
      "13s - loss: 9.1123e-04 - acc: 2.9240e-04 - val_loss: 0.0014 - val_acc: 0.0013\n",
      "Epoch 37/50\n",
      "13s - loss: 0.0010 - acc: 2.9240e-04 - val_loss: 0.0019 - val_acc: 0.0013\n",
      "Epoch 38/50\n",
      "14s - loss: 8.6145e-04 - acc: 2.9240e-04 - val_loss: 0.0098 - val_acc: 0.0013\n",
      "Epoch 39/50\n",
      "14s - loss: 8.4498e-04 - acc: 2.9240e-04 - val_loss: 0.0012 - val_acc: 0.0013\n",
      "Epoch 40/50\n",
      "590s - loss: 8.9179e-04 - acc: 2.9240e-04 - val_loss: 3.3264e-04 - val_acc: 0.0013\n",
      "Epoch 41/50\n",
      "14s - loss: 9.0552e-04 - acc: 2.9240e-04 - val_loss: 4.4956e-04 - val_acc: 0.0013\n",
      "Epoch 42/50\n",
      "14s - loss: 7.7122e-04 - acc: 2.9240e-04 - val_loss: 4.2487e-04 - val_acc: 0.0013\n",
      "Epoch 43/50\n",
      "14s - loss: 8.7359e-04 - acc: 2.9240e-04 - val_loss: 3.4036e-04 - val_acc: 0.0013\n",
      "Epoch 44/50\n",
      "14s - loss: 8.6112e-04 - acc: 2.9240e-04 - val_loss: 0.0026 - val_acc: 0.0013\n",
      "Epoch 45/50\n",
      "14s - loss: 7.5775e-04 - acc: 2.9240e-04 - val_loss: 0.0029 - val_acc: 0.0013\n",
      "Epoch 46/50\n",
      "13s - loss: 7.5955e-04 - acc: 2.9240e-04 - val_loss: 0.0026 - val_acc: 0.0013\n",
      "Epoch 47/50\n",
      "14s - loss: 7.9929e-04 - acc: 2.9240e-04 - val_loss: 7.2207e-04 - val_acc: 0.0013\n",
      "Epoch 48/50\n",
      "13s - loss: 7.2587e-04 - acc: 2.9240e-04 - val_loss: 0.0015 - val_acc: 0.0013\n",
      "Epoch 49/50\n",
      "13s - loss: 7.6243e-04 - acc: 2.9240e-04 - val_loss: 0.0049 - val_acc: 0.0013\n",
      "Epoch 50/50\n",
      "13s - loss: 8.6351e-04 - acc: 2.9240e-04 - val_loss: 2.8649e-04 - val_acc: 0.0013\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1191cc090>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, batch_size=256, nb_epoch=50, validation_data = (X_test, y_test), verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "756/756 [==============================] - 0s     \n"
     ]
    }
   ],
   "source": [
    "score= model.evaluate(X_test, y_test, batch_size=256)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.00028648800860287216, 0.0013227512488368327]\n"
     ]
    }
   ],
   "source": [
    "print score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_pred= model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11921f950>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(y_pred, 'r-')\n",
    "plt.plot(y_test, 'b-')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAELCAYAAAA2mZrgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ng1gszJYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x121c56190>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot (y_pred, y_test, 'bo')\n",
    "plt.xlabel('y_pred')\n",
    "plt.ylabel('y_test')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}