Predict_currency (5).ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import keras.layers.core as core\n",
    "import keras.layers.convolutional as conv\n",
    "import keras.models as models\n",
    "import keras.utils.np_utils as kutils\n",
    "import numpy.random as rnd\n",
    "import matplotlib.pylab as plt\n",
    "import seaborn as sns\n",
    "sns.despine()\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Dropout, Activation, Flatten, ActivityRegularization\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers import Merge\n",
    "from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, CSVLogger, EarlyStopping\n",
    "from keras.optimizers import RMSprop, Adam, SGD, Nadam\n",
    "from keras.layers.advanced_activations import *\n",
    "from keras.layers import Convolution1D, MaxPooling1D, AtrousConvolution1D\n",
    "from keras.layers.recurrent import LSTM, GRU\n",
    "from keras import regularizers\n",
    "\n",
    "import theano\n",
    "theano.config.compute_test_value = \"ignore\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.metrics import mean_squared_error\n",
    "import matplotlib.pylab as plt\n",
    "from pandas import Series\n",
    "from pandas import DataFrame\n",
    "from pandas import concat\n",
    "def shuffle_in_unison(a, b):\n",
    "    # courtsey http://stackoverflow.com/users/190280/josh-bleecher-snyder\n",
    "    assert len(a) == len(b)\n",
    "    shuffled_a = np.empty(a.shape, dtype=a.dtype)\n",
    "    shuffled_b = np.empty(b.shape, dtype=b.dtype)\n",
    "    permutation = np.random.permutation(len(a))\n",
    "    for old_index, new_index in enumerate(permutation):\n",
    "        shuffled_a[new_index] = a[old_index]\n",
    "        shuffled_b[new_index] = b[old_index]\n",
    "    return shuffled_a, shuffled_b\n",
    " \n",
    "def create_Xt_Yt(X, y, percentage=0.9):\n",
    "    p = int(len(X) * percentage)\n",
    "    X = X.values.tolist()\n",
    "    X_train = X[0:p]\n",
    "    Y_train = y[0:p]\n",
    "     \n",
    "    X_train, Y_train = shuffle_in_unison(X_train, Y_train)\n",
    " \n",
    "    X_test = X[p:]\n",
    "    Y_test = y[p:]\n",
    "\n",
    "    return X_train, X_test, Y_train, Y_test\n",
    "# convert an array of values into a dataset matrix\n",
    "def create_dataset(dataset, look_back=1):\n",
    "    dataX, dataY = [], []\n",
    "    for i in range(len(dataset)-look_back-1):\n",
    "        a = dataset[i:(i+look_back)]\n",
    "        dataX.append(a)\n",
    "        if dataset[i + look_back] > dataset[i + look_back - 1]:\n",
    "            dataY.append([0,1])\n",
    "        else:\n",
    "            dataY.append([1,0])\n",
    "    return np.array(dataX), np.array(dataY)\n",
    "\n",
    "def timeseries_to_supervised(data, lag=1,withY  = False):\n",
    "    df = DataFrame(data)\n",
    "    columns = [df.shift(i) for i in range(1, lag+1)]\n",
    "    if withY:\n",
    "        columns.append(df)\n",
    "    df = concat(columns, axis=1)\n",
    "    df.fillna(0, inplace=True)\n",
    "    return df\n",
    "\n",
    "def difference(dataset, interval=1):\n",
    "    diff = list()\n",
    "    for i in range(interval, len(dataset)):\n",
    "        value = dataset[i] - dataset[i - interval]\n",
    "        diff.append(value)\n",
    "    return Series(diff)\n",
    " \n",
    "# invert differenced value\n",
    "def inverse_difference(history, yhat, interval=1):\n",
    "    return yhat + history[-interval]\n",
    " \n",
    "# scale train and test data to [-1, 1]\n",
    "def scale(train, test):\n",
    "    # fit scaler\n",
    "    scaler = MinMaxScaler(feature_range=(-1, 1))\n",
    "    scaler = scaler.fit(train)\n",
    "    # transform train\n",
    "    train = train.reshape(train.shape[0], train.shape[1])\n",
    "    train_scaled = scaler.transform(train)\n",
    "    # transform test\n",
    "    test = test.reshape(test.shape[0], test.shape[1])\n",
    "    test_scaled = scaler.transform(test)\n",
    "    return scaler, train_scaled, test_scaled\n",
    " \n",
    "# inverse scaling for a forecasted value\n",
    "def invert_scale(scaler, X, value):\n",
    "    new_row = [x for x in X] + [yhat]\n",
    "    array = numpy.array(new_row)\n",
    "    array = array.reshape(1, len(array))\n",
    "    inverted = scaler.inverse_transform(array)\n",
    "    return inverted[0, -1]\n",
    "\n",
    "def make_X_LSTM(data, withY = True, look_back = 1):\n",
    "    X = data\n",
    "    if withY == True:\n",
    "        X= data[:, 0:-1]\n",
    "    res = []\n",
    "    for j in range(look_back-1,X.shape[0]):\n",
    "        tmp = []\n",
    "        for i in range(look_back-1,-1,-1):\n",
    "            tmp.append(X[j-i,:])\n",
    "        res.append(tmp)\n",
    "        \n",
    "    #X = X.reshape(X.shape[0], 1, X.shape[1])\n",
    "    return np.array(res)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Unnamed: 0', 'USDRUB.Open', 'USDRUB.High', 'USDRUB.Low',\n",
       "       'USDRUB.Close', 'USDRUB.Volume', 'ClCl.USDRUB.Open',\n",
       "       'RSI.10.USDRUB.Open', 'RSI.20.USDRUB.Open', 'RSI.40.USDRUB.Open',\n",
       "       'SMA.5.USDRUB.Open', 'SMA.10.USDRUB.Open', 'SMA.30.USDRUB.Open',\n",
       "       'SMA.100.USDRUB.Open', 'MACD.nFast.12.nSlow.26.nSig.9.macd.USDRUB.Open',\n",
       "       'MACD.nFast.12.nSlow.26.nSig.9.signal.USDRUB.Open',\n",
       "       'MACD.nFast.2.nSlow.26.nSig.9.macd.USDRUB.Open',\n",
       "       'MACD.nFast.2.nSlow.26.nSig.9.signal.USDRUB.Open',\n",
       "       'MACD.nFast.5.nSlow.26.nSig.9.macd.USDRUB.Open',\n",
       "       'MACD.nFast.5.nSlow.26.nSig.9.signal.USDRUB.Open',\n",
       "       'MACD.nFast.12.nSlow.14.nSig.9.macd.USDRUB.Open',\n",
       "       'MACD.nFast.12.nSlow.14.nSig.9.signal.USDRUB.Open',\n",
       "       'MACD.nFast.2.nSlow.14.nSig.9.macd.USDRUB.Open',\n",
       "       'MACD.nFast.2.nSlow.14.nSig.9.signal.USDRUB.Open',\n",
       "       'MACD.nFast.5.nSlow.14.nSig.9.macd.USDRUB.Open',\n",
       "       'MACD.nFast.5.nSlow.14.nSig.9.signal.USDRUB.Open',\n",
       "       'BBands.n.20.sd.2.dn.USDRUB.Open', 'BBands.n.20.sd.2.mavg.USDRUB.Open',\n",
       "       'BBands.n.20.sd.2.up.USDRUB.Open', 'BBands.n.20.sd.2.pctB.USDRUB.Open',\n",
       "       'BBands.n.50.sd.2.dn.USDRUB.Open', 'BBands.n.50.sd.2.mavg.USDRUB.Open',\n",
       "       'BBands.n.50.sd.2.up.USDRUB.Open', 'BBands.n.50.sd.2.pctB.USDRUB.Open',\n",
       "       'BBands.n.80.sd.2.dn.USDRUB.Open', 'BBands.n.80.sd.2.mavg.USDRUB.Open',\n",
       "       'BBands.n.80.sd.2.up.USDRUB.Open', 'BBands.n.80.sd.2.pctB.USDRUB.Open',\n",
       "       'stoch.nFastK.14.nFastD.1.nSlowD.3.fastK.USDRUB.Open',\n",
       "       'stoch.nFastK.14.nFastD.1.nSlowD.3.fastD.USDRUB.Open',\n",
       "       'stoch.nFastK.14.nFastD.1.nSlowD.3.slowD.USDRUB.Open',\n",
       "       'stoch.nFastK.10.nFastD.1.nSlowD.3.fastK.USDRUB.Open',\n",
       "       'stoch.nFastK.10.nFastD.1.nSlowD.3.fastD.USDRUB.Open',\n",
       "       'stoch.nFastK.10.nFastD.1.nSlowD.3.slowD.USDRUB.Open',\n",
       "       'stoch.nFastK.7.nFastD.1.nSlowD.3.fastK.USDRUB.Open',\n",
       "       'stoch.nFastK.7.nFastD.1.nSlowD.3.fastD.USDRUB.Open',\n",
       "       'stoch.nFastK.7.nFastD.1.nSlowD.3.slowD.USDRUB.Open',\n",
       "       'stoch.nFastK.14.nFastD.3.nSlowD.3.fastK.USDRUB.Open',\n",
       "       'stoch.nFastK.14.nFastD.3.nSlowD.3.fastD.USDRUB.Open',\n",
       "       'stoch.nFastK.14.nFastD.3.nSlowD.3.slowD.USDRUB.Open',\n",
       "       'stoch.nFastK.10.nFastD.3.nSlowD.3.fastK.USDRUB.Open',\n",
       "       'stoch.nFastK.10.nFastD.3.nSlowD.3.fastD.USDRUB.Open',\n",
       "       'stoch.nFastK.10.nFastD.3.nSlowD.3.slowD.USDRUB.Open',\n",
       "       'stoch.nFastK.7.nFastD.3.nSlowD.3.fastK.USDRUB.Open',\n",
       "       'stoch.nFastK.7.nFastD.3.nSlowD.3.fastD.USDRUB.Open',\n",
       "       'stoch.nFastK.7.nFastD.3.nSlowD.3.slowD.USDRUB.Open',\n",
       "       'SMI.n.13.nFast.2.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
       "       'SMI.n.13.nFast.2.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
       "       'SMI.n.10.nFast.2.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
       "       'SMI.n.10.nFast.2.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
       "       'SMI.n.7.nFast.2.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
       "       'SMI.n.7.nFast.2.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
       "       'SMI.n.13.nFast.4.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
       "       'SMI.n.13.nFast.4.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
       "       'SMI.n.10.nFast.4.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
       "       'SMI.n.10.nFast.4.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
       "       'SMI.n.7.nFast.4.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
       "       'SMI.n.7.nFast.4.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
       "       'SMI.n.13.nFast.2.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
       "       'SMI.n.13.nFast.2.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
       "       'SMI.n.10.nFast.2.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
       "       'SMI.n.10.nFast.2.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
       "       'SMI.n.7.nFast.2.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
       "       'SMI.n.7.nFast.2.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
       "       'SMI.n.13.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
       "       'SMI.n.13.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
       "       'SMI.n.10.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
       "       'SMI.n.10.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
       "       'SMI.n.7.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
       "       'SMI.n.7.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
       "       'ZigZag.3.USDRUB.Open', 'ZigZag.5.USDRUB.Open', 'ZigZag.10.USDRUB.Open',\n",
       "       'ZigZag.20.USDRUB.Open'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#read from csv \n",
    "from numpy import genfromtxt\n",
    "import os\n",
    "data = pd.read_csv('data_day1.csv')\n",
    "data.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>USDRUB.Open</th>\n",
       "      <th>USDRUB.High</th>\n",
       "      <th>USDRUB.Low</th>\n",
       "      <th>USDRUB.Close</th>\n",
       "      <th>USDRUB.Volume</th>\n",
       "      <th>ClCl.USDRUB.Open</th>\n",
       "      <th>RSI.10.USDRUB.Open</th>\n",
       "      <th>RSI.20.USDRUB.Open</th>\n",
       "      <th>SMA.5.USDRUB.Open</th>\n",
       "      <th>...</th>\n",
       "      <th>BBands.n.20.sd.2.up.USDRUB.Open</th>\n",
       "      <th>BBands.n.20.sd.2.pctB.USDRUB.Open</th>\n",
       "      <th>BBands.n.50.sd.2.dn.USDRUB.Open</th>\n",
       "      <th>BBands.n.50.sd.2.mavg.USDRUB.Open</th>\n",
       "      <th>BBands.n.50.sd.2.up.USDRUB.Open</th>\n",
       "      <th>BBands.n.50.sd.2.pctB.USDRUB.Open</th>\n",
       "      <th>BBands.n.80.sd.2.dn.USDRUB.Open</th>\n",
       "      <th>BBands.n.80.sd.2.mavg.USDRUB.Open</th>\n",
       "      <th>BBands.n.80.sd.2.up.USDRUB.Open</th>\n",
       "      <th>BBands.n.80.sd.2.pctB.USDRUB.Open</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1999-05-20</td>\n",
       "      <td>24.700</td>\n",
       "      <td>24.700</td>\n",
       "      <td>24.700</td>\n",
       "      <td>24.700</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.002020</td>\n",
       "      <td>46.948452</td>\n",
       "      <td>49.745158</td>\n",
       "      <td>25.0996</td>\n",
       "      <td>...</td>\n",
       "      <td>25.511010</td>\n",
       "      <td>0.509050</td>\n",
       "      <td>23.124503</td>\n",
       "      <td>24.862567</td>\n",
       "      <td>26.600631</td>\n",
       "      <td>0.453233</td>\n",
       "      <td>21.822025</td>\n",
       "      <td>24.144729</td>\n",
       "      <td>26.467433</td>\n",
       "      <td>0.619531</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1999-05-21</td>\n",
       "      <td>24.650</td>\n",
       "      <td>24.650</td>\n",
       "      <td>24.650</td>\n",
       "      <td>24.650</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.002024</td>\n",
       "      <td>46.010874</td>\n",
       "      <td>49.303043</td>\n",
       "      <td>24.8898</td>\n",
       "      <td>...</td>\n",
       "      <td>25.487106</td>\n",
       "      <td>0.487720</td>\n",
       "      <td>23.204641</td>\n",
       "      <td>24.888767</td>\n",
       "      <td>26.572892</td>\n",
       "      <td>0.429113</td>\n",
       "      <td>21.861551</td>\n",
       "      <td>24.167604</td>\n",
       "      <td>26.473657</td>\n",
       "      <td>0.604593</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1999-05-24</td>\n",
       "      <td>24.600</td>\n",
       "      <td>24.600</td>\n",
       "      <td>24.600</td>\n",
       "      <td>24.600</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.002028</td>\n",
       "      <td>45.012088</td>\n",
       "      <td>48.846069</td>\n",
       "      <td>24.6700</td>\n",
       "      <td>...</td>\n",
       "      <td>25.459931</td>\n",
       "      <td>0.467413</td>\n",
       "      <td>23.280795</td>\n",
       "      <td>24.912767</td>\n",
       "      <td>26.544738</td>\n",
       "      <td>0.404175</td>\n",
       "      <td>21.892425</td>\n",
       "      <td>24.186854</td>\n",
       "      <td>26.481284</td>\n",
       "      <td>0.590032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1999-05-25</td>\n",
       "      <td>24.634</td>\n",
       "      <td>24.634</td>\n",
       "      <td>24.634</td>\n",
       "      <td>24.634</td>\n",
       "      <td>0</td>\n",
       "      <td>0.001382</td>\n",
       "      <td>45.899408</td>\n",
       "      <td>49.183209</td>\n",
       "      <td>24.6668</td>\n",
       "      <td>...</td>\n",
       "      <td>25.463672</td>\n",
       "      <td>0.484263</td>\n",
       "      <td>23.361365</td>\n",
       "      <td>24.937447</td>\n",
       "      <td>26.513529</td>\n",
       "      <td>0.403734</td>\n",
       "      <td>21.924503</td>\n",
       "      <td>24.206654</td>\n",
       "      <td>26.488805</td>\n",
       "      <td>0.593628</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1999-05-26</td>\n",
       "      <td>24.634</td>\n",
       "      <td>24.634</td>\n",
       "      <td>24.634</td>\n",
       "      <td>24.634</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>45.899408</td>\n",
       "      <td>49.183209</td>\n",
       "      <td>24.6436</td>\n",
       "      <td>...</td>\n",
       "      <td>25.466954</td>\n",
       "      <td>0.480661</td>\n",
       "      <td>23.445596</td>\n",
       "      <td>24.962127</td>\n",
       "      <td>26.478657</td>\n",
       "      <td>0.391817</td>\n",
       "      <td>21.976931</td>\n",
       "      <td>24.232079</td>\n",
       "      <td>26.487227</td>\n",
       "      <td>0.589112</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 23 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0  USDRUB.Open  USDRUB.High  USDRUB.Low  USDRUB.Close  \\\n",
       "0  1999-05-20       24.700       24.700      24.700        24.700   \n",
       "1  1999-05-21       24.650       24.650      24.650        24.650   \n",
       "2  1999-05-24       24.600       24.600      24.600        24.600   \n",
       "3  1999-05-25       24.634       24.634      24.634        24.634   \n",
       "4  1999-05-26       24.634       24.634      24.634        24.634   \n",
       "\n",
       "   USDRUB.Volume  ClCl.USDRUB.Open  RSI.10.USDRUB.Open  RSI.20.USDRUB.Open  \\\n",
       "0              0         -0.002020           46.948452           49.745158   \n",
       "1              0         -0.002024           46.010874           49.303043   \n",
       "2              0         -0.002028           45.012088           48.846069   \n",
       "3              0          0.001382           45.899408           49.183209   \n",
       "4              0          0.000000           45.899408           49.183209   \n",
       "\n",
       "   SMA.5.USDRUB.Open                ...                  \\\n",
       "0            25.0996                ...                   \n",
       "1            24.8898                ...                   \n",
       "2            24.6700                ...                   \n",
       "3            24.6668                ...                   \n",
       "4            24.6436                ...                   \n",
       "\n",
       "   BBands.n.20.sd.2.up.USDRUB.Open  BBands.n.20.sd.2.pctB.USDRUB.Open  \\\n",
       "0                        25.511010                           0.509050   \n",
       "1                        25.487106                           0.487720   \n",
       "2                        25.459931                           0.467413   \n",
       "3                        25.463672                           0.484263   \n",
       "4                        25.466954                           0.480661   \n",
       "\n",
       "   BBands.n.50.sd.2.dn.USDRUB.Open  BBands.n.50.sd.2.mavg.USDRUB.Open  \\\n",
       "0                        23.124503                          24.862567   \n",
       "1                        23.204641                          24.888767   \n",
       "2                        23.280795                          24.912767   \n",
       "3                        23.361365                          24.937447   \n",
       "4                        23.445596                          24.962127   \n",
       "\n",
       "   BBands.n.50.sd.2.up.USDRUB.Open  BBands.n.50.sd.2.pctB.USDRUB.Open  \\\n",
       "0                        26.600631                           0.453233   \n",
       "1                        26.572892                           0.429113   \n",
       "2                        26.544738                           0.404175   \n",
       "3                        26.513529                           0.403734   \n",
       "4                        26.478657                           0.391817   \n",
       "\n",
       "   BBands.n.80.sd.2.dn.USDRUB.Open  BBands.n.80.sd.2.mavg.USDRUB.Open  \\\n",
       "0                        21.822025                          24.144729   \n",
       "1                        21.861551                          24.167604   \n",
       "2                        21.892425                          24.186854   \n",
       "3                        21.924503                          24.206654   \n",
       "4                        21.976931                          24.232079   \n",
       "\n",
       "   BBands.n.80.sd.2.up.USDRUB.Open  BBands.n.80.sd.2.pctB.USDRUB.Open  \n",
       "0                        26.467433                           0.619531  \n",
       "1                        26.473657                           0.604593  \n",
       "2                        26.481284                           0.590032  \n",
       "3                        26.488805                           0.593628  \n",
       "4                        26.487227                           0.589112  \n",
       "\n",
       "[5 rows x 23 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "df = data.copy()\n",
    "X = df[[\"USDRUB.Close\",'RSI.20.USDRUB.Open','BBands.n.50.sd.2.pctB.USDRUB.Open',\n",
    "      'SMA.10.USDRUB.Open','ClCl.USDRUB.Open']]\n",
    "X = df[['USDRUB.Close','USDRUB.Open', 'USDRUB.High', 'USDRUB.Low',\n",
    "       'USDRUB.Volume', 'ClCl.USDRUB.Open',\n",
    "       'RSI.10.USDRUB.Open', 'RSI.20.USDRUB.Open', 'RSI.40.USDRUB.Open',\n",
    "       'SMA.5.USDRUB.Open', 'SMA.10.USDRUB.Open', 'SMA.30.USDRUB.Open',\n",
    "       'SMA.100.USDRUB.Open', 'MACD.nFast.12.nSlow.26.nSig.9.macd.USDRUB.Open',\n",
    "       'MACD.nFast.12.nSlow.26.nSig.9.signal.USDRUB.Open',\n",
    "       'MACD.nFast.2.nSlow.26.nSig.9.macd.USDRUB.Open',\n",
    "       'MACD.nFast.2.nSlow.26.nSig.9.signal.USDRUB.Open',\n",
    "       'MACD.nFast.5.nSlow.26.nSig.9.macd.USDRUB.Open',\n",
    "       'MACD.nFast.5.nSlow.26.nSig.9.signal.USDRUB.Open',\n",
    "       'MACD.nFast.12.nSlow.14.nSig.9.macd.USDRUB.Open',\n",
    "       'MACD.nFast.12.nSlow.14.nSig.9.signal.USDRUB.Open',\n",
    "       'MACD.nFast.2.nSlow.14.nSig.9.macd.USDRUB.Open',\n",
    "       'MACD.nFast.2.nSlow.14.nSig.9.signal.USDRUB.Open',\n",
    "       'MACD.nFast.5.nSlow.14.nSig.9.macd.USDRUB.Open',\n",
    "       'MACD.nFast.5.nSlow.14.nSig.9.signal.USDRUB.Open',\n",
    "       'BBands.n.20.sd.2.dn.USDRUB.Open', 'BBands.n.20.sd.2.mavg.USDRUB.Open',\n",
    "       'BBands.n.20.sd.2.up.USDRUB.Open', 'BBands.n.20.sd.2.pctB.USDRUB.Open',\n",
    "       'BBands.n.50.sd.2.dn.USDRUB.Open', 'BBands.n.50.sd.2.mavg.USDRUB.Open',\n",
    "       'BBands.n.50.sd.2.up.USDRUB.Open', 'BBands.n.50.sd.2.pctB.USDRUB.Open',\n",
    "       'BBands.n.80.sd.2.dn.USDRUB.Open', 'BBands.n.80.sd.2.mavg.USDRUB.Open',\n",
    "       'BBands.n.80.sd.2.up.USDRUB.Open', 'BBands.n.80.sd.2.pctB.USDRUB.Open',\n",
    "       'stoch.nFastK.14.nFastD.1.nSlowD.3.fastK.USDRUB.Open',\n",
    "       'stoch.nFastK.14.nFastD.1.nSlowD.3.fastD.USDRUB.Open',\n",
    "       'stoch.nFastK.14.nFastD.1.nSlowD.3.slowD.USDRUB.Open',\n",
    "       'stoch.nFastK.10.nFastD.1.nSlowD.3.fastK.USDRUB.Open',\n",
    "       'stoch.nFastK.10.nFastD.1.nSlowD.3.fastD.USDRUB.Open',\n",
    "       'stoch.nFastK.10.nFastD.1.nSlowD.3.slowD.USDRUB.Open',\n",
    "       'stoch.nFastK.7.nFastD.1.nSlowD.3.fastK.USDRUB.Open',\n",
    "       'stoch.nFastK.7.nFastD.1.nSlowD.3.fastD.USDRUB.Open',\n",
    "       'stoch.nFastK.7.nFastD.1.nSlowD.3.slowD.USDRUB.Open',\n",
    "       'stoch.nFastK.14.nFastD.3.nSlowD.3.fastK.USDRUB.Open',\n",
    "       'stoch.nFastK.14.nFastD.3.nSlowD.3.fastD.USDRUB.Open',\n",
    "       'stoch.nFastK.14.nFastD.3.nSlowD.3.slowD.USDRUB.Open',\n",
    "       'stoch.nFastK.10.nFastD.3.nSlowD.3.fastK.USDRUB.Open',\n",
    "       'stoch.nFastK.10.nFastD.3.nSlowD.3.fastD.USDRUB.Open',\n",
    "       'stoch.nFastK.10.nFastD.3.nSlowD.3.slowD.USDRUB.Open',\n",
    "       'stoch.nFastK.7.nFastD.3.nSlowD.3.fastK.USDRUB.Open',\n",
    "       'stoch.nFastK.7.nFastD.3.nSlowD.3.fastD.USDRUB.Open',\n",
    "       'stoch.nFastK.7.nFastD.3.nSlowD.3.slowD.USDRUB.Open',\n",
    "       'SMI.n.13.nFast.2.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
    "       'SMI.n.13.nFast.2.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
    "       'SMI.n.10.nFast.2.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
    "       'SMI.n.10.nFast.2.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
    "       'SMI.n.7.nFast.2.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
    "       'SMI.n.7.nFast.2.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
    "       'SMI.n.13.nFast.4.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
    "       'SMI.n.13.nFast.4.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
    "       'SMI.n.10.nFast.4.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
    "       'SMI.n.10.nFast.4.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
    "       'SMI.n.7.nFast.4.nSlow.25.nSig.9.SMI.USDRUB.Open',\n",
    "       'SMI.n.7.nFast.4.nSlow.25.nSig.9.signal.USDRUB.Open',\n",
    "       'SMI.n.13.nFast.2.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
    "       'SMI.n.13.nFast.2.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
    "       'SMI.n.10.nFast.2.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
    "       'SMI.n.10.nFast.2.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
    "       'SMI.n.7.nFast.2.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
    "       'SMI.n.7.nFast.2.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
    "       'SMI.n.13.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
    "       'SMI.n.13.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
    "       'SMI.n.10.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
    "       'SMI.n.10.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
    "       'SMI.n.7.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open',\n",
    "       'SMI.n.7.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open',\n",
    "       'ZigZag.3.USDRUB.Open', 'ZigZag.5.USDRUB.Open', 'ZigZag.10.USDRUB.Open',\n",
    "       'ZigZag.20.USDRUB.Open']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>USDRUB.Close</th>\n",
       "      <th>USDRUB.Open</th>\n",
       "      <th>USDRUB.High</th>\n",
       "      <th>USDRUB.Low</th>\n",
       "      <th>USDRUB.Volume</th>\n",
       "      <th>ClCl.USDRUB.Open</th>\n",
       "      <th>RSI.10.USDRUB.Open</th>\n",
       "      <th>RSI.20.USDRUB.Open</th>\n",
       "      <th>RSI.40.USDRUB.Open</th>\n",
       "      <th>SMA.5.USDRUB.Open</th>\n",
       "      <th>...</th>\n",
       "      <th>SMI.n.13.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open</th>\n",
       "      <th>SMI.n.13.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open</th>\n",
       "      <th>SMI.n.10.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open</th>\n",
       "      <th>SMI.n.10.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open</th>\n",
       "      <th>SMI.n.7.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open</th>\n",
       "      <th>SMI.n.7.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open</th>\n",
       "      <th>ZigZag.3.USDRUB.Open</th>\n",
       "      <th>ZigZag.5.USDRUB.Open</th>\n",
       "      <th>ZigZag.10.USDRUB.Open</th>\n",
       "      <th>ZigZag.20.USDRUB.Open</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>24.445</td>\n",
       "      <td>24.494</td>\n",
       "      <td>24.494</td>\n",
       "      <td>24.445</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.002652</td>\n",
       "      <td>44.123574</td>\n",
       "      <td>47.410102</td>\n",
       "      <td>50.612298</td>\n",
       "      <td>24.4806</td>\n",
       "      <td>...</td>\n",
       "      <td>-36.012539</td>\n",
       "      <td>-36.930992</td>\n",
       "      <td>-28.319462</td>\n",
       "      <td>-29.528078</td>\n",
       "      <td>-19.624278</td>\n",
       "      <td>-20.963279</td>\n",
       "      <td>25.054566</td>\n",
       "      <td>25.054566</td>\n",
       "      <td>24.501317</td>\n",
       "      <td>23.753317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>24.464</td>\n",
       "      <td>24.464</td>\n",
       "      <td>24.464</td>\n",
       "      <td>24.464</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000777</td>\n",
       "      <td>45.306581</td>\n",
       "      <td>47.759057</td>\n",
       "      <td>50.729166</td>\n",
       "      <td>24.5006</td>\n",
       "      <td>...</td>\n",
       "      <td>-35.418205</td>\n",
       "      <td>-36.325877</td>\n",
       "      <td>-27.724099</td>\n",
       "      <td>-28.806486</td>\n",
       "      <td>-19.155829</td>\n",
       "      <td>-20.240299</td>\n",
       "      <td>25.026547</td>\n",
       "      <td>25.026547</td>\n",
       "      <td>24.509926</td>\n",
       "      <td>23.762744</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>24.464</td>\n",
       "      <td>24.464</td>\n",
       "      <td>24.464</td>\n",
       "      <td>24.464</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>45.306581</td>\n",
       "      <td>47.759057</td>\n",
       "      <td>50.729166</td>\n",
       "      <td>24.4736</td>\n",
       "      <td>...</td>\n",
       "      <td>-34.850001</td>\n",
       "      <td>-35.735527</td>\n",
       "      <td>-27.198877</td>\n",
       "      <td>-28.163443</td>\n",
       "      <td>-18.804624</td>\n",
       "      <td>-19.666029</td>\n",
       "      <td>24.998528</td>\n",
       "      <td>24.998528</td>\n",
       "      <td>24.518535</td>\n",
       "      <td>23.772172</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>24.434</td>\n",
       "      <td>24.464</td>\n",
       "      <td>24.464</td>\n",
       "      <td>24.434</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.001226</td>\n",
       "      <td>43.510859</td>\n",
       "      <td>47.210994</td>\n",
       "      <td>50.530562</td>\n",
       "      <td>24.4634</td>\n",
       "      <td>...</td>\n",
       "      <td>-34.556633</td>\n",
       "      <td>-35.263969</td>\n",
       "      <td>-27.053656</td>\n",
       "      <td>-27.719528</td>\n",
       "      <td>-18.967840</td>\n",
       "      <td>-19.386753</td>\n",
       "      <td>24.970509</td>\n",
       "      <td>24.970509</td>\n",
       "      <td>24.527144</td>\n",
       "      <td>23.781599</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>24.344</td>\n",
       "      <td>24.434</td>\n",
       "      <td>24.434</td>\n",
       "      <td>24.344</td>\n",
       "      <td>0</td>\n",
       "      <td>-0.003683</td>\n",
       "      <td>38.433205</td>\n",
       "      <td>45.559961</td>\n",
       "      <td>49.929110</td>\n",
       "      <td>24.4302</td>\n",
       "      <td>...</td>\n",
       "      <td>-35.146372</td>\n",
       "      <td>-35.216930</td>\n",
       "      <td>-28.056912</td>\n",
       "      <td>-27.854481</td>\n",
       "      <td>-20.631140</td>\n",
       "      <td>-19.884508</td>\n",
       "      <td>24.942491</td>\n",
       "      <td>24.942491</td>\n",
       "      <td>24.535753</td>\n",
       "      <td>23.791027</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 83 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   USDRUB.Close  USDRUB.Open  USDRUB.High  USDRUB.Low  USDRUB.Volume  \\\n",
       "0        24.445       24.494       24.494      24.445              0   \n",
       "1        24.464       24.464       24.464      24.464              0   \n",
       "2        24.464       24.464       24.464      24.464              0   \n",
       "3        24.434       24.464       24.464      24.434              0   \n",
       "4        24.344       24.434       24.434      24.344              0   \n",
       "\n",
       "   ClCl.USDRUB.Open  RSI.10.USDRUB.Open  RSI.20.USDRUB.Open  \\\n",
       "0         -0.002652           44.123574           47.410102   \n",
       "1          0.000777           45.306581           47.759057   \n",
       "2          0.000000           45.306581           47.759057   \n",
       "3         -0.001226           43.510859           47.210994   \n",
       "4         -0.003683           38.433205           45.559961   \n",
       "\n",
       "   RSI.40.USDRUB.Open  SMA.5.USDRUB.Open          ...            \\\n",
       "0           50.612298            24.4806          ...             \n",
       "1           50.729166            24.5006          ...             \n",
       "2           50.729166            24.4736          ...             \n",
       "3           50.530562            24.4634          ...             \n",
       "4           49.929110            24.4302          ...             \n",
       "\n",
       "   SMI.n.13.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open  \\\n",
       "0                                        -36.012539   \n",
       "1                                        -35.418205   \n",
       "2                                        -34.850001   \n",
       "3                                        -34.556633   \n",
       "4                                        -35.146372   \n",
       "\n",
       "   SMI.n.13.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open  \\\n",
       "0                                         -36.930992     \n",
       "1                                         -36.325877     \n",
       "2                                         -35.735527     \n",
       "3                                         -35.263969     \n",
       "4                                         -35.216930     \n",
       "\n",
       "   SMI.n.10.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open  \\\n",
       "0                                        -28.319462   \n",
       "1                                        -27.724099   \n",
       "2                                        -27.198877   \n",
       "3                                        -27.053656   \n",
       "4                                        -28.056912   \n",
       "\n",
       "   SMI.n.10.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open  \\\n",
       "0                                         -29.528078     \n",
       "1                                         -28.806486     \n",
       "2                                         -28.163443     \n",
       "3                                         -27.719528     \n",
       "4                                         -27.854481     \n",
       "\n",
       "   SMI.n.7.nFast.4.nSlow.25.nSig.4.SMI.USDRUB.Open  \\\n",
       "0                                       -19.624278   \n",
       "1                                       -19.155829   \n",
       "2                                       -18.804624   \n",
       "3                                       -18.967840   \n",
       "4                                       -20.631140   \n",
       "\n",
       "   SMI.n.7.nFast.4.nSlow.25.nSig.4.signal.USDRUB.Open  ZigZag.3.USDRUB.Open  \\\n",
       "0                                         -20.963279              25.054566   \n",
       "1                                         -20.240299              25.026547   \n",
       "2                                         -19.666029              24.998528   \n",
       "3                                         -19.386753              24.970509   \n",
       "4                                         -19.884508              24.942491   \n",
       "\n",
       "   ZigZag.5.USDRUB.Open  ZigZag.10.USDRUB.Open  ZigZag.20.USDRUB.Open  \n",
       "0             25.054566              24.501317              23.753317  \n",
       "1             25.026547              24.509926              23.762744  \n",
       "2             24.998528              24.518535              23.772172  \n",
       "3             24.970509              24.527144              23.781599  \n",
       "4             24.942491              24.535753              23.791027  \n",
       "\n",
       "[5 rows x 83 columns]"
      ]
     },
     "execution_count": 122,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# not to change look_back!!!\n",
    "look_back = 1\n",
    "res = DataFrame()\n",
    "for i in range(X.shape[1]-1,-1,-1):\n",
    "    raw_values = X[[i]].values\n",
    "    raw_values = raw_values.reshape(raw_values.shape[0])\n",
    "    diff_values = difference(raw_values, 1)\n",
    "    if i == 0:\n",
    "        supervised = timeseries_to_supervised(diff_values, look_back,withY = True)\n",
    "    else:\n",
    "        supervised = timeseries_to_supervised(diff_values, look_back,withY = False)\n",
    "    res = pd.concat([res,DataFrame(supervised.values)],axis = 1)\n",
    "supervised_values = res.values\n",
    "# split data into train and test-sets\n",
    "p = 0.7\n",
    "train, test = supervised_values[0:int(len(supervised_values) * p)], supervised_values[int(len(supervised_values) * p):]\n",
    " "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "batch_size = 20\n",
    "look_back = 1\n",
    "scaler, trainX, testX = scale(train[:,:-1],\n",
    "                              test[:,:-1])\n",
    "trainX = make_X_LSTM(trainX, withY = False,look_back = look_back)\n",
    "testX = make_X_LSTM(testX, withY = False,look_back = look_back)\n",
    "\n",
    "trainX = trainX[0:(trainX.shape[0]//batch_size) * batch_size,:,:]\n",
    "testX = testX[0:(testX.shape[0]//batch_size) * batch_size,:,:]\n",
    "\n",
    "#train Y\n",
    "trainY = train[(look_back - 1):,-1] > 0\n",
    "trainY = trainY[0:(trainY.shape[0]//batch_size) * batch_size,]\n",
    "#test Y\n",
    "testY = test[(look_back - 1):,-1] > 0\n",
    "testY = testY[0:(testY.shape[0]//batch_size) * batch_size,]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_56 (LSTM)               (20, 1, 2)                688       \n",
      "_________________________________________________________________\n",
      "activity_regularization_26 ( (20, 1, 2)                0         \n",
      "_________________________________________________________________\n",
      "dropout_55 (Dropout)         (20, 1, 2)                0         \n",
      "_________________________________________________________________\n",
      "lstm_57 (LSTM)               (20, 8)                   352       \n",
      "_________________________________________________________________\n",
      "dropout_56 (Dropout)         (20, 8)                   0         \n",
      "_________________________________________________________________\n",
      "dense_55 (Dense)             (20, 8)                   72        \n",
      "_________________________________________________________________\n",
      "activation_55 (Activation)   (20, 8)                   0         \n",
      "_________________________________________________________________\n",
      "dense_56 (Dense)             (20, 1)                   9         \n",
      "_________________________________________________________________\n",
      "activation_56 (Activation)   (20, 1)                   0         \n",
      "=================================================================\n",
      "Total params: 1,121.0\n",
      "Trainable params: 1,121.0\n",
      "Non-trainable params: 0.0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "model = Sequential()\n",
    "model.add(LSTM(2, batch_input_shape=(batch_size, trainX.shape[1], trainX.shape[2]), \n",
    "               stateful=True,\n",
    "               return_sequences=True))\n",
    "model.add(ActivityRegularization(l1=0.0001, l2=0.0001))\n",
    "model.add(Dropout(0.15))\n",
    "model.add(LSTM(8))\n",
    "model.add(Dropout(0.15))\n",
    "model.add(Dense(8))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dense(1))\n",
    "model.add(Activation('sigmoid'))\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=\"test.hdf5\", verbose=1, save_best_only=True)\n",
    "model.compile(loss='binary_crossentropy',\n",
    "          optimizer='adam',\n",
    "          class_mode=\"binary\",\n",
    "          metrics=['accuracy'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "best_acc = 0\n",
    "lowest_loss = 10000000\n",
    "train_acc = []\n",
    "test_acc = []\n",
    "train_loss = []\n",
    "test_loss = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Block 1 / 50\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/keras/models.py:826: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n",
      "  warnings.warn('The `nb_epoch` argument in `fit` '\n",
      "/usr/local/lib/python3.5/dist-packages/keras/backend/tensorflow_backend.py:2094: UserWarning: Expected no kwargs, you passed 1\n",
      "kwargs passed to function are ignored with Tensorflow backend\n",
      "  warnings.warn('\\n'.join(msg))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train acc: 0.5644 ,     Test acc: 0.4817\n",
      "Model successfully saved  val_acc =  0.4817\n",
      "Block 2 / 50\n",
      "Train acc: 0.5647 ,     Test acc: 0.4817\n",
      "Block 3 / 50\n",
      "Train acc: 0.5647 ,     Test acc: 0.4817\n",
      "Block 4 / 50\n",
      "Train acc: 0.5647 ,     Test acc: 0.4817\n",
      "Block 5 / 50\n",
      "Train acc: 0.5647 ,     Test acc: 0.4817\n",
      "Model successfully saved  val_acc =  0.4817\n",
      "Block 6 / 50\n",
      "Train acc: 0.5649 ,     Test acc: 0.4835\n",
      "Model successfully saved  val_acc =  0.4835\n",
      "Block 7 / 50\n",
      "Train acc: 0.5662 ,     Test acc: 0.4909\n",
      "Model successfully saved  val_acc =  0.4909\n",
      "Block 8 / 50\n",
      "Train acc: 0.5686 ,     Test acc: 0.5195\n",
      "Model successfully saved  val_acc =  0.5195\n",
      "Block 9 / 50\n",
      "Train acc: 0.5725 ,     Test acc: 0.5384\n",
      "Model successfully saved  val_acc =  0.5384\n",
      "Block 10 / 50\n",
      "Train acc: 0.5783 ,     Test acc: 0.5506\n",
      "Model successfully saved  val_acc =  0.5506\n",
      "Block 11 / 50\n",
      "Train acc: 0.5764 ,     Test acc: 0.5579\n",
      "Block 12 / 50\n",
      "Train acc: 0.5819 ,     Test acc: 0.5500\n",
      "Block 13 / 50\n",
      "Train acc: 0.5796 ,     Test acc: 0.5585\n",
      "Block 14 / 50\n",
      "Train acc: 0.5843 ,     Test acc: 0.5543\n",
      "Block 15 / 50\n",
      "Train acc: 0.5843 ,     Test acc: 0.5530\n",
      "Block 16 / 50\n",
      "Train acc: 0.5869 ,     Test acc: 0.5494\n",
      "Block 17 / 50\n",
      "Train acc: 0.5856 ,     Test acc: 0.5470\n",
      "Block 18 / 50\n",
      "Train acc: 0.5906 ,     Test acc: 0.5463\n",
      "Block 19 / 50\n",
      "Train acc: 0.5848 ,     Test acc: 0.5445\n",
      "Block 20 / 50\n",
      "Train acc: 0.5924 ,     Test acc: 0.5445\n",
      "Block 21 / 50\n",
      "Train acc: 0.5921 ,     Test acc: 0.5409\n",
      "Block 22 / 50\n",
      "Train acc: 0.5908 ,     Test acc: 0.5378\n",
      "Block 23 / 50\n",
      "Train acc: 0.5880 ,     Test acc: 0.5293\n",
      "Block 24 / 50\n",
      "Train acc: 0.5903 ,     Test acc: 0.5287\n",
      "Block 25 / 50\n",
      "Train acc: 0.5916 ,     Test acc: 0.5293\n",
      "Block 26 / 50\n",
      "Train acc: 0.5958 ,     Test acc: 0.5323\n",
      "Block 27 / 50\n",
      "Train acc: 0.5911 ,     Test acc: 0.5329\n",
      "Block 28 / 50\n",
      "Train acc: 0.5950 ,     Test acc: 0.5311\n",
      "Block 29 / 50\n",
      "Train acc: 0.5948 ,     Test acc: 0.5329\n",
      "Block 30 / 50\n",
      "Train acc: 0.5908 ,     Test acc: 0.5317\n",
      "Block 31 / 50\n",
      "Train acc: 0.5932 ,     Test acc: 0.5262\n",
      "Block 32 / 50\n",
      "Train acc: 0.6000 ,     Test acc: 0.5244\n",
      "Block 33 / 50\n",
      "Train acc: 0.6026 ,     Test acc: 0.5280\n",
      "Block 34 / 50\n",
      "Train acc: 0.5921 ,     Test acc: 0.5256\n",
      "Block 35 / 50\n",
      "Train acc: 0.5874 ,     Test acc: 0.5220\n",
      "Block 36 / 50\n",
      "Train acc: 0.6029 ,     Test acc: 0.5244\n",
      "Block 37 / 50\n",
      "Train acc: 0.5979 ,     Test acc: 0.5274\n",
      "Block 38 / 50\n",
      "Train acc: 0.6003 ,     Test acc: 0.5220\n",
      "Block 39 / 50\n",
      "Train acc: 0.5995 ,     Test acc: 0.5238\n",
      "Block 40 / 50\n",
      "Train acc: 0.5984 ,     Test acc: 0.5256\n",
      "Block 41 / 50\n",
      "Train acc: 0.6034 ,     Test acc: 0.5262\n",
      "Block 42 / 50\n",
      "Train acc: 0.5974 ,     Test acc: 0.5305\n",
      "Block 43 / 50\n",
      "Train acc: 0.6000 ,     Test acc: 0.5256\n",
      "Block 44 / 50\n",
      "Train acc: 0.5969 ,     Test acc: 0.5256\n",
      "Block 45 / 50\n",
      "Train acc: 0.6013 ,     Test acc: 0.5207\n",
      "Block 46 / 50\n",
      "Train acc: 0.5937 ,     Test acc: 0.5201\n",
      "Block 47 / 50\n",
      "Train acc: 0.5958 ,     Test acc: 0.5159\n",
      "Block 48 / 50\n",
      "Train acc: 0.6003 ,     Test acc: 0.5140\n",
      "Block 49 / 50\n",
      "Train acc: 0.5948 ,     Test acc: 0.5195\n",
      "Block 50 / 50\n",
      "Train acc: 0.5987 ,     Test acc: 0.5201\n"
     ]
    }
   ],
   "source": [
    "\n",
    "nchar = 4\n",
    "nb_epoch = 50\n",
    "for i in range(nb_epoch):\n",
    "    print(\"Block\",i + 1,\"/\",nb_epoch)\n",
    "    history = model.fit(trainX, \n",
    "                        trainY, \n",
    "                        nb_epoch=1,\n",
    "                        batch_size=batch_size, \n",
    "                        verbose=0, \n",
    "                        shuffle=False, \n",
    "                        validation_data=(testX, testY)\n",
    "         )\n",
    "    print(\"Train acc:\",'%.4f' % round(history.history['acc'][0], nchar) ,\",    \",\n",
    "         \"Test acc:\", '%.4f' % round(history.history['val_acc'][0], nchar))\n",
    "    train_acc.append(history.history['acc'][0])\n",
    "    train_loss.append(history.history['loss'][0])\n",
    "    test_acc.append(history.history['val_acc'][0])\n",
    "    test_loss.append(history.history['val_loss'][0])\n",
    "    if history.history['val_loss'][0] < lowest_loss:\n",
    "        model.save('best_model.h5')\n",
    "        best_acc = history.history['val_acc'][0]\n",
    "        lowest_loss = history.history['val_loss'][0]\n",
    "        print(\"Model successfully saved \",\"val_acc = \",'%.4f' % round(best_acc, nchar))\n",
    "    model.reset_states()\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 240,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.55060975726057848"
      ]
     },
     "execution_count": 240,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_acc\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfUAAAFnCAYAAAC/5tBZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niroQQghhJ6SoCyGEEHZCiroQQghhJ/4fv5gsGxSTtqgAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb584bab3c8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfUAAAFnCAYAAAC/5tBZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nvPr2oVcJdQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb586335be0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "plt.plot(train_loss)\n",
    "plt.plot(test_loss)\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='best')\n",
    "plt.show()\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(train_acc)\n",
    "plt.plot(test_acc)\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='best')\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": {
  "anaconda-cloud": {},
  "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}