Yangzijiang EPOCH100 LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\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",
       "      <td>3174</td>\n",
       "      <td>12/16/2019</td>\n",
       "      <td>1.13</td>\n",
       "      <td>1.13</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.10</td>\n",
       "      <td>1.10</td>\n",
       "      <td>27908700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3175</td>\n",
       "      <td>12/17/2019</td>\n",
       "      <td>1.10</td>\n",
       "      <td>1.11</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.07</td>\n",
       "      <td>32804600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3176</td>\n",
       "      <td>12/18/2019</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.11</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.09</td>\n",
       "      <td>38128700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3177</td>\n",
       "      <td>12/19/2019</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.08</td>\n",
       "      <td>20343100</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3178</td>\n",
       "      <td>12/20/2019</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.07</td>\n",
       "      <td>33596100</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Date  Open  High   Low  Close  Adj Close    Volume\n",
       "3174  12/16/2019  1.13  1.13  1.09   1.10       1.10  27908700\n",
       "3175  12/17/2019  1.10  1.11  1.07   1.07       1.07  32804600\n",
       "3176  12/18/2019  1.08  1.11  1.08   1.09       1.09  38128700\n",
       "3177  12/19/2019  1.09  1.09  1.07   1.08       1.08  20343100\n",
       "3178  12/20/2019  1.08  1.09  1.07   1.07       1.07  33596100"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd \n",
    "\n",
    "\n",
    "df_Yzj = pd.read_csv('BS6.SI.CSV')\n",
    "df_Yzj.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\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",
       "      <td>0</td>\n",
       "      <td>4/18/2007</td>\n",
       "      <td>1.35</td>\n",
       "      <td>1.39</td>\n",
       "      <td>1.33</td>\n",
       "      <td>1.34</td>\n",
       "      <td>0.842029</td>\n",
       "      <td>413736000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>4/19/2007</td>\n",
       "      <td>1.35</td>\n",
       "      <td>1.35</td>\n",
       "      <td>1.26</td>\n",
       "      <td>1.27</td>\n",
       "      <td>0.798042</td>\n",
       "      <td>95804</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>4/20/2007</td>\n",
       "      <td>1.28</td>\n",
       "      <td>1.31</td>\n",
       "      <td>1.28</td>\n",
       "      <td>1.30</td>\n",
       "      <td>0.816894</td>\n",
       "      <td>70064</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>4/23/2007</td>\n",
       "      <td>1.31</td>\n",
       "      <td>1.31</td>\n",
       "      <td>1.26</td>\n",
       "      <td>1.27</td>\n",
       "      <td>0.798042</td>\n",
       "      <td>36099</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4/24/2007</td>\n",
       "      <td>1.26</td>\n",
       "      <td>1.27</td>\n",
       "      <td>1.26</td>\n",
       "      <td>1.27</td>\n",
       "      <td>0.798042</td>\n",
       "      <td>14422</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3174</td>\n",
       "      <td>12/16/2019</td>\n",
       "      <td>1.13</td>\n",
       "      <td>1.13</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.10</td>\n",
       "      <td>1.100000</td>\n",
       "      <td>27908700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3175</td>\n",
       "      <td>12/17/2019</td>\n",
       "      <td>1.10</td>\n",
       "      <td>1.11</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.070000</td>\n",
       "      <td>32804600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3176</td>\n",
       "      <td>12/18/2019</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.11</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.090000</td>\n",
       "      <td>38128700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3177</td>\n",
       "      <td>12/19/2019</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.080000</td>\n",
       "      <td>20343100</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3178</td>\n",
       "      <td>12/20/2019</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.070000</td>\n",
       "      <td>33596100</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3179 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            Date  Open  High   Low  Close  Adj Close     Volume\n",
       "0      4/18/2007  1.35  1.39  1.33   1.34   0.842029  413736000\n",
       "1      4/19/2007  1.35  1.35  1.26   1.27   0.798042      95804\n",
       "2      4/20/2007  1.28  1.31  1.28   1.30   0.816894      70064\n",
       "3      4/23/2007  1.31  1.31  1.26   1.27   0.798042      36099\n",
       "4      4/24/2007  1.26  1.27  1.26   1.27   0.798042      14422\n",
       "...          ...   ...   ...   ...    ...        ...        ...\n",
       "3174  12/16/2019  1.13  1.13  1.09   1.10   1.100000   27908700\n",
       "3175  12/17/2019  1.10  1.11  1.07   1.07   1.070000   32804600\n",
       "3176  12/18/2019  1.08  1.11  1.08   1.09   1.090000   38128700\n",
       "3177  12/19/2019  1.09  1.09  1.07   1.08   1.080000   20343100\n",
       "3178  12/20/2019  1.08  1.09  1.07   1.07   1.070000   33596100\n",
       "\n",
       "[3179 rows x 7 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_Yzj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "plt.figure()\n",
    "plt.plot(df_Yzj[\"Open\"])\n",
    "plt.plot(df_Yzj[\"High\"])\n",
    "plt.plot(df_Yzj[\"Low\"])\n",
    "plt.plot(df_Yzj[\"Close\"])\n",
    "plt.title('Yangzijiang Stock price history')\n",
    "plt.ylabel('Price (SGD)')\n",
    "plt.xlabel('Days')\n",
    "plt.legend(['Open','High','Low','Close'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "plt.plot(df_Yzj[\"Volume\"])\n",
    "plt.title('Yangzijiang stock volume history')\n",
    "plt.ylabel('Volume')\n",
    "plt.xlabel('Days')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "checking if any null values are present\n",
      " Date         0\n",
      "Open         0\n",
      "High         0\n",
      "Low          0\n",
      "Close        0\n",
      "Adj Close    0\n",
      "Volume       0\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print(\"checking if any null values are present\\n\", df_Yzj.isna().sum())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_Yzj = df_Yzj.dropna() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "checking if any null values are present\n",
      " Date         0\n",
      "Open         0\n",
      "High         0\n",
      "Low          0\n",
      "Close        0\n",
      "Adj Close    0\n",
      "Volume       0\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print(\"checking if any null values are present\\n\", df_Yzj.isna().sum())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\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",
       "      <td>0</td>\n",
       "      <td>4/18/2007</td>\n",
       "      <td>1.35</td>\n",
       "      <td>1.39</td>\n",
       "      <td>1.33</td>\n",
       "      <td>1.34</td>\n",
       "      <td>0.842029</td>\n",
       "      <td>413736000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>4/19/2007</td>\n",
       "      <td>1.35</td>\n",
       "      <td>1.35</td>\n",
       "      <td>1.26</td>\n",
       "      <td>1.27</td>\n",
       "      <td>0.798042</td>\n",
       "      <td>95804</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>4/20/2007</td>\n",
       "      <td>1.28</td>\n",
       "      <td>1.31</td>\n",
       "      <td>1.28</td>\n",
       "      <td>1.30</td>\n",
       "      <td>0.816894</td>\n",
       "      <td>70064</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>4/23/2007</td>\n",
       "      <td>1.31</td>\n",
       "      <td>1.31</td>\n",
       "      <td>1.26</td>\n",
       "      <td>1.27</td>\n",
       "      <td>0.798042</td>\n",
       "      <td>36099</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4/24/2007</td>\n",
       "      <td>1.26</td>\n",
       "      <td>1.27</td>\n",
       "      <td>1.26</td>\n",
       "      <td>1.27</td>\n",
       "      <td>0.798042</td>\n",
       "      <td>14422</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3174</td>\n",
       "      <td>12/16/2019</td>\n",
       "      <td>1.13</td>\n",
       "      <td>1.13</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.10</td>\n",
       "      <td>1.100000</td>\n",
       "      <td>27908700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3175</td>\n",
       "      <td>12/17/2019</td>\n",
       "      <td>1.10</td>\n",
       "      <td>1.11</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.070000</td>\n",
       "      <td>32804600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3176</td>\n",
       "      <td>12/18/2019</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.11</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.090000</td>\n",
       "      <td>38128700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3177</td>\n",
       "      <td>12/19/2019</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.080000</td>\n",
       "      <td>20343100</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3178</td>\n",
       "      <td>12/20/2019</td>\n",
       "      <td>1.08</td>\n",
       "      <td>1.09</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.07</td>\n",
       "      <td>1.070000</td>\n",
       "      <td>33596100</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3179 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            Date  Open  High   Low  Close  Adj Close     Volume\n",
       "0      4/18/2007  1.35  1.39  1.33   1.34   0.842029  413736000\n",
       "1      4/19/2007  1.35  1.35  1.26   1.27   0.798042      95804\n",
       "2      4/20/2007  1.28  1.31  1.28   1.30   0.816894      70064\n",
       "3      4/23/2007  1.31  1.31  1.26   1.27   0.798042      36099\n",
       "4      4/24/2007  1.26  1.27  1.26   1.27   0.798042      14422\n",
       "...          ...   ...   ...   ...    ...        ...        ...\n",
       "3174  12/16/2019  1.13  1.13  1.09   1.10   1.100000   27908700\n",
       "3175  12/17/2019  1.10  1.11  1.07   1.07   1.070000   32804600\n",
       "3176  12/18/2019  1.08  1.11  1.08   1.09   1.090000   38128700\n",
       "3177  12/19/2019  1.09  1.09  1.07   1.08   1.080000   20343100\n",
       "3178  12/20/2019  1.08  1.09  1.07   1.07   1.070000   33596100\n",
       "\n",
       "[3179 rows x 7 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_Yzj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train and Test size 2543 636\n"
     ]
    }
   ],
   "source": [
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "train_cols = [\"Open\",\"High\",\"Low\",\"Close\",\"Volume\"]\n",
    "df_train, df_test = train_test_split(df_Yzj, train_size=0.8, test_size=0.2, shuffle=False)\n",
    "print(\"Train and Test size\", len(df_train), len(df_test))\n",
    "# scale the feature MinMax, build array\n",
    "x = df_train.loc[:,train_cols].values\n",
    "min_max_scaler = MinMaxScaler()\n",
    "x_train = min_max_scaler.fit_transform(x)\n",
    "x_test = min_max_scaler.transform(df_test.loc[:,train_cols])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_timeseries(mat, y_col_index):\n",
    "    # y_col_index is the index of column that would act as output column\n",
    "    # total number of time-series samples would be len(mat) - TIME_STEPS\n",
    "    dim_0 = mat.shape[0] - TIME_STEPS\n",
    "    dim_1 = mat.shape[1]\n",
    "    x = np.zeros((dim_0, TIME_STEPS, dim_1))\n",
    "    y = np.zeros((dim_0,))\n",
    "    \n",
    "    for i in tqdm_notebook(range(dim_0)):\n",
    "        x[i] = mat[i:TIME_STEPS+i]\n",
    "        y[i] = mat[TIME_STEPS+i, y_col_index]\n",
    "    print(\"length of time-series i/o\",x.shape,y.shape)\n",
    "    return x, y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def trim_dataset(mat, batch_size):\n",
    "    \"\"\"\n",
    "    trims dataset to a size that's divisible by BATCH_SIZE\n",
    "    \"\"\"\n",
    "    no_of_rows_drop = mat.shape[0]%batch_size\n",
    "    if(no_of_rows_drop > 0):\n",
    "        return mat[:-no_of_rows_drop]\n",
    "    else:\n",
    "        return mat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "import numpy as np\n",
    "from tqdm._tqdm_notebook import tqdm_notebook\n",
    "params = {\n",
    "    \"batch_size\": 20,  # 20<16<10, 25 was a bust\n",
    "    ##\"epochs\": 300,\n",
    "    \"epochs\":100,\n",
    "    \"lr\": 0.00010000,\n",
    "    \"time_steps\": 60\n",
    "}\n",
    "\n",
    "TIME_STEPS = params[\"time_steps\"]\n",
    "BATCH_SIZE = params[\"batch_size\"]\n",
    "stime = time.time()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "72dde0c0edc143bf881c23de7d7535dc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=2483), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "length of time-series i/o (2483, 60, 5) (2483,)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "28b92e9e30d947ee9def549ee36e2791",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=576), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "length of time-series i/o (576, 60, 5) (576,)\n"
     ]
    }
   ],
   "source": [
    "x_t, y_t = build_timeseries(x_train, 3)\n",
    "x_t = trim_dataset(x_t, BATCH_SIZE)\n",
    "y_t = trim_dataset(y_t, BATCH_SIZE)\n",
    "x_temp, y_temp = build_timeseries(x_test, 3)\n",
    "x_val, x_test_t = np.split(trim_dataset(x_temp, BATCH_SIZE),2)\n",
    "y_val, y_test_t = np.split(trim_dataset(y_temp, BATCH_SIZE),2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential, load_model\n",
    "from keras.layers import Dense, Dropout\n",
    "from keras.layers import LSTM\n",
    "from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau, CSVLogger\n",
    "from keras import optimizers\n",
    "\n",
    "\n",
    "def create_model():\n",
    "    lstm_model = Sequential()\n",
    "    lstm_model.add(LSTM(100, batch_input_shape=(BATCH_SIZE, TIME_STEPS, x_t.shape[2]), dropout=0.0, recurrent_dropout=0.0, stateful=True, kernel_initializer='random_uniform'))\n",
    "    lstm_model.add(Dropout(0.5))\n",
    "    lstm_model.add(Dense(20,activation='relu'))\n",
    "    lstm_model.add(Dense(1,activation='sigmoid'))\n",
    "    optimizer = optimizers.RMSprop(lr=params[\"lr\"])\n",
    "    lstm_model.compile(loss='mean_squared_error', optimizer=optimizer)\n",
    "    lstm_model.summary()\n",
    "    return lstm_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'C:\\\\Users\\\\Victor\\\\Desktop\\\\project5\\\\Stock-Price-Prediction\\\\outputyzjLSTM'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import os\n",
    "#OUTPUT_PATH ='/home/victor/Stock-Price-Prediction/outputs/'\n",
    "OUTPUT_PATH ='C:\\\\Users\\\\Victor\\\\Desktop\\\\project5\\\\Stock-Price-Prediction\\\\outputyzjLSTM'\n",
    "OUTPUT_PATH"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'C:\\\\Users\\\\Victor\\\\Desktop\\\\project5\\\\Stock-Price-Prediction'"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model not found\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5cf7dab860124629b27dd8dcd7b1fe34",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=576), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "length of time-series i/o (576, 60, 5) (576,)\n",
      "Test size (280, 60, 5) (280,) (280, 60, 5) (280,)\n",
      "Building model...\n",
      "checking if GPU available []\n",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_1 (LSTM)                (20, 100)                 42400     \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (20, 100)                 0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (20, 20)                  2020      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (20, 1)                   21        \n",
      "=================================================================\n",
      "Total params: 44,441\n",
      "Trainable params: 44,441\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Train on 2480 samples, validate on 280 samples\n",
      "Epoch 1/100\n",
      " - 6s - loss: 0.0393 - val_loss: 0.0401\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.04010, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 2/100\n",
      " - 5s - loss: 0.0374 - val_loss: 0.0355\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.04010 to 0.03554, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 3/100\n",
      " - 5s - loss: 0.0326 - val_loss: 0.0233\n",
      "\n",
      "Epoch 00003: val_loss improved from 0.03554 to 0.02330, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 4/100\n",
      " - 5s - loss: 0.0243 - val_loss: 0.0107\n",
      "\n",
      "Epoch 00004: val_loss improved from 0.02330 to 0.01072, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 5/100\n",
      " - 5s - loss: 0.0125 - val_loss: 0.0063\n",
      "\n",
      "Epoch 00005: val_loss improved from 0.01072 to 0.00627, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 6/100\n",
      " - 5s - loss: 0.0086 - val_loss: 0.0055\n",
      "\n",
      "Epoch 00006: val_loss improved from 0.00627 to 0.00551, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 7/100\n",
      " - 5s - loss: 0.0074 - val_loss: 0.0050\n",
      "\n",
      "Epoch 00007: val_loss improved from 0.00551 to 0.00495, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 8/100\n",
      " - 5s - loss: 0.0070 - val_loss: 0.0046\n",
      "\n",
      "Epoch 00008: val_loss improved from 0.00495 to 0.00464, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 9/100\n",
      " - 5s - loss: 0.0065 - val_loss: 0.0043\n",
      "\n",
      "Epoch 00009: val_loss improved from 0.00464 to 0.00431, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 10/100\n",
      " - 5s - loss: 0.0060 - val_loss: 0.0044\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.00431\n",
      "Epoch 11/100\n",
      " - 5s - loss: 0.0056 - val_loss: 0.0040\n",
      "\n",
      "Epoch 00011: val_loss improved from 0.00431 to 0.00401, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 12/100\n",
      " - 5s - loss: 0.0053 - val_loss: 0.0039\n",
      "\n",
      "Epoch 00012: val_loss improved from 0.00401 to 0.00393, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 13/100\n",
      " - 5s - loss: 0.0048 - val_loss: 0.0038\n",
      "\n",
      "Epoch 00013: val_loss improved from 0.00393 to 0.00382, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 14/100\n",
      " - 5s - loss: 0.0046 - val_loss: 0.0037\n",
      "\n",
      "Epoch 00014: val_loss improved from 0.00382 to 0.00372, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 15/100\n",
      " - 5s - loss: 0.0043 - val_loss: 0.0034\n",
      "\n",
      "Epoch 00015: val_loss improved from 0.00372 to 0.00343, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 16/100\n",
      " - 5s - loss: 0.0040 - val_loss: 0.0037\n",
      "\n",
      "Epoch 00016: val_loss did not improve from 0.00343\n",
      "Epoch 17/100\n",
      " - 5s - loss: 0.0036 - val_loss: 0.0033\n",
      "\n",
      "Epoch 00017: val_loss improved from 0.00343 to 0.00330, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 18/100\n",
      " - 5s - loss: 0.0035 - val_loss: 0.0030\n",
      "\n",
      "Epoch 00018: val_loss improved from 0.00330 to 0.00305, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 19/100\n",
      " - 5s - loss: 0.0031 - val_loss: 0.0030\n",
      "\n",
      "Epoch 00019: val_loss improved from 0.00305 to 0.00296, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 20/100\n",
      " - 5s - loss: 0.0031 - val_loss: 0.0027\n",
      "\n",
      "Epoch 00020: val_loss improved from 0.00296 to 0.00273, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 21/100\n",
      " - 5s - loss: 0.0027 - val_loss: 0.0028\n",
      "\n",
      "Epoch 00021: val_loss did not improve from 0.00273\n",
      "Epoch 22/100\n",
      " - 5s - loss: 0.0027 - val_loss: 0.0023\n",
      "\n",
      "Epoch 00022: val_loss improved from 0.00273 to 0.00235, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 23/100\n",
      " - 5s - loss: 0.0026 - val_loss: 0.0024\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.00235\n",
      "Epoch 24/100\n",
      " - 5s - loss: 0.0026 - val_loss: 0.0024\n",
      "\n",
      "Epoch 00024: val_loss did not improve from 0.00235\n",
      "Epoch 25/100\n",
      " - 5s - loss: 0.0025 - val_loss: 0.0020\n",
      "\n",
      "Epoch 00025: val_loss improved from 0.00235 to 0.00196, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 26/100\n",
      " - 5s - loss: 0.0024 - val_loss: 0.0021\n",
      "\n",
      "Epoch 00026: val_loss did not improve from 0.00196\n",
      "Epoch 27/100\n",
      " - 5s - loss: 0.0022 - val_loss: 0.0020\n",
      "\n",
      "Epoch 00027: val_loss did not improve from 0.00196\n",
      "Epoch 28/100\n",
      " - 5s - loss: 0.0022 - val_loss: 0.0019\n",
      "\n",
      "Epoch 00028: val_loss improved from 0.00196 to 0.00194, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 29/100\n",
      " - 5s - loss: 0.0022 - val_loss: 0.0019\n",
      "\n",
      "Epoch 00029: val_loss improved from 0.00194 to 0.00187, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 30/100\n",
      " - 6s - loss: 0.0021 - val_loss: 0.0018\n",
      "\n",
      "Epoch 00030: val_loss improved from 0.00187 to 0.00183, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 31/100\n",
      " - 5s - loss: 0.0021 - val_loss: 0.0015\n",
      "\n",
      "Epoch 00031: val_loss improved from 0.00183 to 0.00151, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 32/100\n",
      " - 5s - loss: 0.0021 - val_loss: 0.0021\n",
      "\n",
      "Epoch 00032: val_loss did not improve from 0.00151\n",
      "Epoch 33/100\n",
      " - 5s - loss: 0.0021 - val_loss: 0.0015\n",
      "\n",
      "Epoch 00033: val_loss did not improve from 0.00151\n",
      "Epoch 34/100\n",
      " - 5s - loss: 0.0021 - val_loss: 0.0017\n",
      "\n",
      "Epoch 00034: val_loss did not improve from 0.00151\n",
      "Epoch 35/100\n",
      " - 5s - loss: 0.0020 - val_loss: 0.0014\n",
      "\n",
      "Epoch 00035: val_loss improved from 0.00151 to 0.00140, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 36/100\n",
      " - 5s - loss: 0.0021 - val_loss: 0.0017\n",
      "\n",
      "Epoch 00036: val_loss did not improve from 0.00140\n",
      "Epoch 37/100\n",
      " - 5s - loss: 0.0021 - val_loss: 0.0017\n",
      "\n",
      "Epoch 00037: val_loss did not improve from 0.00140\n",
      "Epoch 38/100\n",
      " - 5s - loss: 0.0021 - val_loss: 0.0015\n",
      "\n",
      "Epoch 00038: val_loss did not improve from 0.00140\n",
      "Epoch 39/100\n",
      " - 5s - loss: 0.0020 - val_loss: 0.0012\n",
      "\n",
      "Epoch 00039: val_loss improved from 0.00140 to 0.00123, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 40/100\n",
      " - 5s - loss: 0.0019 - val_loss: 0.0014\n",
      "\n",
      "Epoch 00040: val_loss did not improve from 0.00123\n",
      "Epoch 41/100\n",
      " - 5s - loss: 0.0018 - val_loss: 0.0013\n",
      "\n",
      "Epoch 00041: val_loss did not improve from 0.00123\n",
      "Epoch 42/100\n",
      " - 5s - loss: 0.0019 - val_loss: 0.0014\n",
      "\n",
      "Epoch 00042: val_loss did not improve from 0.00123\n",
      "Epoch 43/100\n",
      " - 5s - loss: 0.0019 - val_loss: 0.0012\n",
      "\n",
      "Epoch 00043: val_loss did not improve from 0.00123\n",
      "Epoch 44/100\n",
      " - 5s - loss: 0.0018 - val_loss: 0.0012\n",
      "\n",
      "Epoch 00044: val_loss did not improve from 0.00123\n",
      "Epoch 45/100\n",
      " - 5s - loss: 0.0018 - val_loss: 0.0013\n",
      "\n",
      "Epoch 00045: val_loss did not improve from 0.00123\n",
      "Epoch 46/100\n",
      " - 5s - loss: 0.0018 - val_loss: 0.0013\n",
      "\n",
      "Epoch 00046: val_loss did not improve from 0.00123\n",
      "Epoch 47/100\n",
      " - 5s - loss: 0.0018 - val_loss: 0.0013\n",
      "\n",
      "Epoch 00047: val_loss did not improve from 0.00123\n",
      "Epoch 48/100\n",
      " - 5s - loss: 0.0019 - val_loss: 0.0011\n",
      "\n",
      "Epoch 00048: val_loss improved from 0.00123 to 0.00110, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 49/100\n",
      " - 5s - loss: 0.0018 - val_loss: 0.0014\n",
      "\n",
      "Epoch 00049: val_loss did not improve from 0.00110\n",
      "Epoch 50/100\n",
      " - 5s - loss: 0.0018 - val_loss: 0.0014\n",
      "\n",
      "Epoch 00050: val_loss did not improve from 0.00110\n",
      "Epoch 51/100\n",
      " - 5s - loss: 0.0017 - val_loss: 0.0015\n",
      "\n",
      "Epoch 00051: val_loss did not improve from 0.00110\n",
      "Epoch 52/100\n",
      " - 5s - loss: 0.0017 - val_loss: 0.0012\n",
      "\n",
      "Epoch 00052: val_loss did not improve from 0.00110\n",
      "Epoch 53/100\n",
      " - 5s - loss: 0.0017 - val_loss: 0.0011\n",
      "\n",
      "Epoch 00053: val_loss improved from 0.00110 to 0.00107, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 54/100\n",
      " - 5s - loss: 0.0017 - val_loss: 0.0012\n",
      "\n",
      "Epoch 00054: val_loss did not improve from 0.00107\n",
      "Epoch 55/100\n",
      " - 5s - loss: 0.0017 - val_loss: 0.0012\n",
      "\n",
      "Epoch 00055: val_loss did not improve from 0.00107\n",
      "Epoch 56/100\n",
      " - 5s - loss: 0.0016 - val_loss: 9.7567e-04\n",
      "\n",
      "Epoch 00056: val_loss improved from 0.00107 to 0.00098, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 57/100\n",
      " - 5s - loss: 0.0017 - val_loss: 0.0010\n",
      "\n",
      "Epoch 00057: val_loss did not improve from 0.00098\n",
      "Epoch 58/100\n",
      " - 5s - loss: 0.0017 - val_loss: 0.0010\n",
      "\n",
      "Epoch 00058: val_loss did not improve from 0.00098\n",
      "Epoch 59/100\n",
      " - 5s - loss: 0.0016 - val_loss: 9.2566e-04\n",
      "\n",
      "Epoch 00059: val_loss improved from 0.00098 to 0.00093, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 60/100\n",
      " - 5s - loss: 0.0016 - val_loss: 0.0011\n",
      "\n",
      "Epoch 00060: val_loss did not improve from 0.00093\n",
      "Epoch 61/100\n",
      " - 5s - loss: 0.0016 - val_loss: 9.0936e-04\n",
      "\n",
      "Epoch 00061: val_loss improved from 0.00093 to 0.00091, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 62/100\n",
      " - 5s - loss: 0.0015 - val_loss: 0.0012\n",
      "\n",
      "Epoch 00062: val_loss did not improve from 0.00091\n",
      "Epoch 63/100\n",
      " - 5s - loss: 0.0015 - val_loss: 0.0010\n",
      "\n",
      "Epoch 00063: val_loss did not improve from 0.00091\n",
      "Epoch 64/100\n",
      " - 5s - loss: 0.0016 - val_loss: 9.5193e-04\n",
      "\n",
      "Epoch 00064: val_loss did not improve from 0.00091\n",
      "Epoch 65/100\n",
      " - 5s - loss: 0.0016 - val_loss: 7.8445e-04\n",
      "\n",
      "Epoch 00065: val_loss improved from 0.00091 to 0.00078, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 66/100\n",
      " - 5s - loss: 0.0016 - val_loss: 9.6115e-04\n",
      "\n",
      "Epoch 00066: val_loss did not improve from 0.00078\n",
      "Epoch 67/100\n",
      " - 5s - loss: 0.0015 - val_loss: 9.6171e-04\n",
      "\n",
      "Epoch 00067: val_loss did not improve from 0.00078\n",
      "Epoch 68/100\n",
      " - 5s - loss: 0.0015 - val_loss: 9.3865e-04\n",
      "\n",
      "Epoch 00068: val_loss did not improve from 0.00078\n",
      "Epoch 69/100\n",
      " - 5s - loss: 0.0015 - val_loss: 8.9901e-04\n",
      "\n",
      "Epoch 00069: val_loss did not improve from 0.00078\n",
      "Epoch 70/100\n",
      " - 5s - loss: 0.0015 - val_loss: 7.8950e-04\n",
      "\n",
      "Epoch 00070: val_loss did not improve from 0.00078\n",
      "Epoch 71/100\n",
      " - 5s - loss: 0.0014 - val_loss: 7.6200e-04\n",
      "\n",
      "Epoch 00071: val_loss improved from 0.00078 to 0.00076, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 72/100\n",
      " - 5s - loss: 0.0013 - val_loss: 8.2651e-04\n",
      "\n",
      "Epoch 00072: val_loss did not improve from 0.00076\n",
      "Epoch 73/100\n",
      " - 5s - loss: 0.0014 - val_loss: 7.7028e-04\n",
      "\n",
      "Epoch 00073: val_loss did not improve from 0.00076\n",
      "Epoch 74/100\n",
      " - 5s - loss: 0.0014 - val_loss: 8.2621e-04\n",
      "\n",
      "Epoch 00074: val_loss did not improve from 0.00076\n",
      "Epoch 75/100\n",
      " - 5s - loss: 0.0014 - val_loss: 8.0115e-04\n",
      "\n",
      "Epoch 00075: val_loss did not improve from 0.00076\n",
      "Epoch 76/100\n",
      " - 5s - loss: 0.0014 - val_loss: 8.8521e-04\n",
      "\n",
      "Epoch 00076: val_loss did not improve from 0.00076\n",
      "Epoch 77/100\n",
      " - 5s - loss: 0.0013 - val_loss: 6.9810e-04\n",
      "\n",
      "Epoch 00077: val_loss improved from 0.00076 to 0.00070, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 78/100\n",
      " - 5s - loss: 0.0013 - val_loss: 7.6931e-04\n",
      "\n",
      "Epoch 00078: val_loss did not improve from 0.00070\n",
      "Epoch 79/100\n",
      " - 5s - loss: 0.0013 - val_loss: 7.6662e-04\n",
      "\n",
      "Epoch 00079: val_loss did not improve from 0.00070\n",
      "Epoch 80/100\n",
      " - 5s - loss: 0.0013 - val_loss: 8.0841e-04\n",
      "\n",
      "Epoch 00080: val_loss did not improve from 0.00070\n",
      "Epoch 81/100\n",
      " - 5s - loss: 0.0013 - val_loss: 8.3270e-04\n",
      "\n",
      "Epoch 00081: val_loss did not improve from 0.00070\n",
      "Epoch 82/100\n",
      " - 5s - loss: 0.0014 - val_loss: 8.8060e-04\n",
      "\n",
      "Epoch 00082: val_loss did not improve from 0.00070\n",
      "Epoch 83/100\n",
      " - 5s - loss: 0.0014 - val_loss: 7.0666e-04\n",
      "\n",
      "Epoch 00083: val_loss did not improve from 0.00070\n",
      "Epoch 84/100\n",
      " - 5s - loss: 0.0013 - val_loss: 7.4846e-04\n",
      "\n",
      "Epoch 00084: val_loss did not improve from 0.00070\n",
      "Epoch 85/100\n",
      " - 5s - loss: 0.0013 - val_loss: 7.6934e-04\n",
      "\n",
      "Epoch 00085: val_loss did not improve from 0.00070\n",
      "Epoch 86/100\n",
      " - 5s - loss: 0.0012 - val_loss: 7.1382e-04\n",
      "\n",
      "Epoch 00086: val_loss did not improve from 0.00070\n",
      "Epoch 87/100\n",
      " - 5s - loss: 0.0012 - val_loss: 7.7161e-04\n",
      "\n",
      "Epoch 00087: val_loss did not improve from 0.00070\n",
      "Epoch 88/100\n",
      " - 5s - loss: 0.0012 - val_loss: 6.7246e-04\n",
      "\n",
      "Epoch 00088: val_loss improved from 0.00070 to 0.00067, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 89/100\n",
      " - 7s - loss: 0.0012 - val_loss: 6.3660e-04\n",
      "\n",
      "Epoch 00089: val_loss improved from 0.00067 to 0.00064, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 90/100\n",
      " - 5s - loss: 0.0012 - val_loss: 6.7934e-04\n",
      "\n",
      "Epoch 00090: val_loss did not improve from 0.00064\n",
      "Epoch 91/100\n",
      " - 5s - loss: 0.0013 - val_loss: 6.2553e-04\n",
      "\n",
      "Epoch 00091: val_loss improved from 0.00064 to 0.00063, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 92/100\n",
      " - 6s - loss: 0.0012 - val_loss: 6.8069e-04\n",
      "\n",
      "Epoch 00092: val_loss did not improve from 0.00063\n",
      "Epoch 93/100\n",
      " - 6s - loss: 0.0012 - val_loss: 6.2772e-04\n",
      "\n",
      "Epoch 00093: val_loss did not improve from 0.00063\n",
      "Epoch 94/100\n",
      " - 5s - loss: 0.0011 - val_loss: 6.3854e-04\n",
      "\n",
      "Epoch 00094: val_loss did not improve from 0.00063\n",
      "Epoch 95/100\n",
      " - 5s - loss: 0.0011 - val_loss: 6.8128e-04\n",
      "\n",
      "Epoch 00095: val_loss did not improve from 0.00063\n",
      "Epoch 96/100\n",
      " - 5s - loss: 0.0012 - val_loss: 6.1193e-04\n",
      "\n",
      "Epoch 00096: val_loss improved from 0.00063 to 0.00061, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 97/100\n",
      " - 5s - loss: 0.0011 - val_loss: 6.0392e-04\n",
      "\n",
      "Epoch 00097: val_loss improved from 0.00061 to 0.00060, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 98/100\n",
      " - 5s - loss: 0.0011 - val_loss: 6.0857e-04\n",
      "\n",
      "Epoch 00098: val_loss did not improve from 0.00060\n",
      "Epoch 99/100\n",
      " - 5s - loss: 0.0011 - val_loss: 5.7652e-04\n",
      "\n",
      "Epoch 00099: val_loss improved from 0.00060 to 0.00058, saving model to C:\\Users\\Victor\\Desktop\\project5\\Stock-Price-Prediction\\outputyzjLSTM\\best_model.h5\n",
      "Epoch 100/100\n",
      " - 5s - loss: 0.0010 - val_loss: 6.0443e-04\n",
      "\n",
      "Epoch 00100: val_loss did not improve from 0.00058\n",
      "saving model...\n"
     ]
    }
   ],
   "source": [
    "model = None\n",
    "try:\n",
    "    model = pickle.load(open(\"lstm_model\", 'rb'))\n",
    "    print(\"Loaded saved model...\")\n",
    "except FileNotFoundError:\n",
    "    print(\"Model not found\")\n",
    "\n",
    "\n",
    "x_temp, y_temp = build_timeseries(x_test, 3)\n",
    "x_val, x_test_t = np.split(trim_dataset(x_temp, BATCH_SIZE),2)\n",
    "y_val, y_test_t = np.split(trim_dataset(y_temp, BATCH_SIZE),2)\n",
    "\n",
    "print(\"Test size\", x_test_t.shape, y_test_t.shape, x_val.shape, y_val.shape)\n",
    "    \n",
    "is_update_model = True\n",
    "if model is None or is_update_model:\n",
    "    from keras import backend as K\n",
    "    print(\"Building model...\")\n",
    "    print(\"checking if GPU available\", K.tensorflow_backend._get_available_gpus())\n",
    "    model = create_model()\n",
    "    \n",
    "    es = EarlyStopping(monitor='val_loss', mode='min', verbose=1,\n",
    "                       patience=40, min_delta=0.0001)\n",
    "    \n",
    "    mcp = ModelCheckpoint(os.path.join(OUTPUT_PATH,\n",
    "                          \"best_model.h5\"), monitor='val_loss', verbose=1,\n",
    "                          save_best_only=True, save_weights_only=False, mode='min', period=1)\n",
    "\n",
    "    # Not used here. But leaving it here as a reminder for future\n",
    "    r_lr_plat = ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=30, \n",
    "                                  verbose=0, mode='auto', min_delta=0.0001, cooldown=0, min_lr=0)\n",
    "    \n",
    "   ## csv_logger = CSVLogger(os.path.join(OUTPUT_PATH, 'training_log_' + time.ctime().replace(\" \",\"_\") + '.log'), append=True)\n",
    "    csv_logger = CSVLogger(os.path.join(OUTPUT_PATH, 'training_log_' + time.strftime(\"%a %b %d %H %M %S %Y\", time.localtime()) + '.log'), append=True)\n",
    "    \n",
    "  #  csv_logger = CSVLogger(os.path.join(OUTPUT_PATH, 'training_log_'  + '.log'), append=True)\n",
    "    \n",
    "    history = model.fit(x_t, y_t, epochs=params[\"epochs\"], verbose=2, batch_size=BATCH_SIZE,\n",
    "                        shuffle=False, validation_data=(trim_dataset(x_val, BATCH_SIZE),\n",
    "                        trim_dataset(y_val, BATCH_SIZE)), callbacks=[es, mcp, csv_logger])\n",
    "    \n",
    "    print(\"saving model...\")\n",
    "    pickle.dump(model, open(\"lstm_model\", \"wb\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Square Error is 0.0006281695684143017 (280,) (280,)\n",
      "[0.39329696 0.39159906 0.39057818 0.39178967 0.3956828  0.39218748\n",
      " 0.3856528  0.38021168 0.37496516 0.3665927  0.35923505 0.35427856\n",
      " 0.35155138 0.35104686 0.35337836]\n",
      "[0.3877551  0.39591837 0.41632653 0.44081633 0.36326531 0.37142857\n",
      " 0.3755102  0.37142857 0.36326531 0.35102041 0.3755102  0.36734694\n",
      " 0.3877551  0.4        0.40816327]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import mean_squared_error\n",
    "# model.evaluate(x_test_t, y_test_t, batch_size=BATCH_SIZE\n",
    "y_pred = model.predict(trim_dataset(x_test_t, BATCH_SIZE), batch_size=BATCH_SIZE)\n",
    "y_pred = y_pred.flatten()\n",
    "y_test_t = trim_dataset(y_test_t, BATCH_SIZE)\n",
    "error = mean_squared_error(y_test_t, y_pred)\n",
    "print(\"Mean Square Error is\", error, y_pred.shape, y_test_t.shape)\n",
    "print(y_pred[0:15])\n",
    "print(y_test_t[0:15])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.2535776 1.2494177 1.2469165 1.2498847 1.2594229 1.2508594 1.2348495\n",
      " 1.2215186 1.2086647 1.1881522 1.1701258 1.1579825 1.1513009 1.1500648\n",
      " 1.155777 ]\n",
      "[1.24 1.26 1.31 1.37 1.18 1.2  1.21 1.2  1.18 1.15 1.21 1.19 1.24 1.27\n",
      " 1.29]\n"
     ]
    }
   ],
   "source": [
    "# convert the predicted value to range of real data\n",
    "y_pred_org = (y_pred * min_max_scaler.data_range_[3]) + min_max_scaler.data_min_[3]\n",
    "# min_max_scaler.inverse_transform(y_pred)\n",
    "y_test_t_org = (y_test_t * min_max_scaler.data_range_[3]) + min_max_scaler.data_min_[3]\n",
    "# min_max_scaler.inverse_transform(y_test_t)\n",
    "print(y_pred_org[0:15])\n",
    "print(y_test_t_org[0:15])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "def mean_absolute_percentage_error(y_test_t, y_pred): \n",
    "    y_test_t, y_pred = np.array(y_test_t), np.array(y_pred)\n",
    "    return np.mean(np.abs((y_test_t - y_pred) / y_test_t)) * 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4.701691517719284"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_absolute_percentage_error(y_test_t, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "#score, acc = model.evaluate(trim_dataset(x_test_t,BATCH_SIZE),trim_dataset(y_test_t,BATCH_SIZE),batch_size=BATCH_SIZE)\n",
    "#print('Test score:', score)\n",
    "#print('Test accuracy:', acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAm0AAAFNCAYAAABST1gVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nOzde5xcVZ3v/c+vbl3V10o6nYRcIOGeCxBCwAsoIIrgBVBxAHW84XBAHWfOjPMMo8dRGT0PnNc8isw4zuAIIqOiB0UjBpFRFLwhiYRAwi1CMJ1A0rn0vavr9nv+2Ls7le7q7nTS1dWdfN+vV72qau+1117VMsOXtfZay9wdEREREZnaItVugIiIiIiMTaFNREREZBpQaBMRERGZBhTaRERERKYBhTYRERGRaUChTURERGQaUGgTEakQM/uMmf3XAZb9hZl96FDrEZHDl0KbiFSFmW0xs9ePcO4TZvaCmXWbWauZfSc8vjE81m1mBTPLlHz/hJm938zczL4wpL7LwuNfH+F+54Xnvz/k+Gnh8V9MzK8WETl4Cm0iMqWY2fuAPwde7+71wCrgZwDuvszd68PjDwMfHfju7v87rOKPwBVmFiup9r3As2Pcug14tZk1lxx73wFcJyIyKRTaRGSqORO4393/CODuL7v7reO4/mXgCeCNAGY2E3g1sHqM67LAD4Arw+uiwJ8B3ywtZGavNrNHzawjfH91ybnFZvZLM+sysweAWUOufaWZ/cbM2s3scTM7bxy/q7SeS8Jex/ZwWHVJybm/N7NtYRueMbMLwuNnmdlaM+s0sx1DeyNFZOpTaBORqeZ3wHvN7O/MbFUYnsbrGwS9axCEsB8C/eO87o3ARmD7wMkwAP4YuAVoBr4A/Likd+5bwDqCsPZPBD11A9fOD6/9HDAT+DjwPTNrGc8PM7MTgW8Dfw20AGuAH5lZwsxOAj4KnOnuDeFv2BJe+iXgS+7eCBwHfHc89xWR6lNoE5Epxd3/C/hLgsDxS2CnmV0/zmruAc4zsyaCEPaNA7z3b4CZYfgpd92bgefc/U53z7v7t4Gngbea2dEEvYSfcvd+d38I+FHJte8B1rj7GncvuvsDwFrgTeP8bVcAP3b3B9w9B/wzkCLoTSwANcBSM4u7+5aBHksgBxxvZrPcvdvdfzfO+4pIlSm0iciU4+7fdPfXA2ngWuAGM3vjOK7vI+jV+l/ALHf/9ThufydBb9X5BOGv1DzgxSHHXgTmh+f2unvPkHMDjgHeGQ5ptptZO3AOcNQ42jasDe5eBLYC8919M0EP3GcIwu5dZjYvLHo1cCLwdDis+5Zx3ldEqkyhTUSmLHfPufv/BTYAy8d5+TeAvyUIYeNxJ/Bhgl6x3iHnthOEr1JHA9uAl4AZZlY35NyArcCd7p4uedW5+43jbN9+bTAzAxaGbcDdv+Xu54RlHLgpPP6cu18FzA6P3T2krSIyxSm0iUg1xc0sWfKKhct2vNnMGswsYmYXA8uAR8ZZ9y+BNwD/Mp6L3P0F4Fzgk2VOrwFONLN3hW29AlgK3OvuLxIMd342fL7sHOCtJdf+F8Ew6hvNLBr+3vPMbME4f9d3gTeb2QVmFicIpv3Ab8zsJDN7nZnVABmgj2DIFDN7j5m1hD1z7WFdhXHeW0SqKDZ2ERGRilkz5PvngT8AnyAIOVGCocDr3P1X46nY3Z1wqZDxGule7r47HFb8EvAVYDPwFnffFRZ5F3AHsAf4LUFvXzq8dquZXQr8H4KJBAXg98B142zbM2b2HoIwOh9YD7zV3bNhWLsRWELwDNtvgGvCSy8CvmBmtQR/0yvdPTOee4tIdVnw/9dEREREZCrT8KiIiIjINKDQJiIiIjINKLSJiIiITAMKbSIiIiLTgEKbiIiIyDRwRCz5MWvWLF+0aFG1myEiIiIypnXr1u1y92H7Eh8RoW3RokWsXbu22s0QERERGZOZDd0uD9DwqIiIiMi0oNAmIiIiMg0otImIiIhMA0fEM23l5HI5WltbyWS09d5ESSaTLFiwgHg8Xu2miIiIHHaO2NDW2tpKQ0MDixYtwsyq3Zxpz93ZvXs3ra2tLF68uNrNEREROewcscOjmUyG5uZmBbYJYmY0Nzer51JERKRCjtjQBiiwTTD9PUVERCqnoqHNzC4ys2fMbLOZXV/mfI2ZfSc8/4iZLRpy/mgz6zazjx9ondPF7t27WbFiBStWrGDu3LnMnz9/8Hs2mz2gOj7wgQ/wzDPPVLilIiIiMhVU7Jk2M4sCXwbeALQCj5rZanffVFLsamCvux9vZlcCNwFXlJz/InDfOOucFpqbm1m/fj0An/nMZ6ivr+fjH//4fmXcHXcnEimfrW+//faKt1NERESmhkr2tJ0FbHb35909C9wFXDqkzKXAHeHnu4ELLBxjM7PLgOeBjeOsc/L1d0Hf3gmpavPmzSxfvpxrr72WlStX8tJLL3HNNdewatUqli1bxg033DBY9pxzzmH9+vXk83nS6TTXX389p512Gq961avYuXPnhLRHREREpoZKhrb5wNaS763hsbJl3D0PdADNZlYH/D3w2YOoc/L1tEHnSxNW3aZNm7j66qt57LHHmD9/PjfeeCNr167l8ccf54EHHmDTpuEdix0dHZx77rk8/vjjvOpVr+K2226bsPaIiIhI9VVyyY9yT6X7AZb5LPBFd+8e8nD7gdQZFDS7BrgG4Oijjx61oZ/90UY2be8ctcyoCtnglWgfbOLSeY18+q3LDqq64447jjPPPHPw+7e//W2+9rWvkc/n2b59O5s2bWLp0qX7XZNKpbj44osBOOOMM3j44YcP7reIiIjIlFTJ0NYKLCz5vgDYPkKZVjOLAU3AHuAVwOVm9n+ANFA0swyw7gDqBMDdbwVuBVi1alXZYDdhLOyw9CJY9JCrq6urG/z83HPP8aUvfYnf//73pNNp3vOe95RdViORSAx+jkaj5PP5Q26HiIiITB2VDG2PAieY2WJgG3Al8K4hZVYD7wN+C1wO/NzdHXjNQAEz+wzQ7e7/Gga7seoct4PtERuU74edm6BpIdTNOtTm7Kezs5OGhgYaGxt56aWXuP/++7nooosm9B4iIiIy9VUstLl73sw+CtwPRIHb3H2jmd0ArHX31cDXgDvNbDNBD9uVB1NnpX7DgerIGg1EiOQnfmHZlStXsnTpUpYvX86xxx7L2WefPeH3EBERkanPgo6tw9uqVat87dq1+x176qmnWLJkyYTUv3VPL819W0jVxLFZJ0xIndPVRP5dRUREjkRmts7dVw09fkTviDBRmusS9JHAc31wBIRgERERmXwKbRMglYhSiCaJeAEv5KrdHBERETkMKbRNADMjlaoHINPXXeXWiIiIyOFIoW2C1NWHoa23p8otERERkcORQtsEiURj5C2O5fvI5gvVbo6IiIgcZhTaJlAkniJJjt092Wo3RURERA4zCm0TKJJIUWNZ2nv6KRZHn0V63nnncf/99+937Oabb+bDH/7wiNfUh0Ow27dv5/LLLx+x3qHLmwx1880309vbO/j9TW96E+3t7aNeIyIiItWl0DaR4ikMiBaztPeNPov0qquu4q677trv2F133cVVV1015m3mzZvH3XfffdDNHBra1qxZQzqdPuj6REREpPIU2iZSLAVAQzTP7u5+Rlu4+PLLL+fee++lv78fgC1btrB9+3ZWrFjBBRdcwMqVKznllFP44Q9/OOzaLVu2sHz5cgD6+vq48sorOfXUU7niiivo6+sbLHfdddexatUqli1bxqc//WkAbrnlFrZv387555/P+eefD8CiRYvYtWsXAF/4whdYvnw5y5cv5+abbx6835IlS/iLv/gLli1bxoUXXrjffURERKTyFNomUqwGMNLxPH25Ar3ZkSckNDc3c9ZZZ/GTn/wECHrZrrjiClKpFPfccw9/+MMfePDBB/nbv/3bUcPfV77yFWpra9mwYQOf/OQnWbdu3eC5z3/+86xdu5YNGzbwy1/+kg0bNvCxj32MefPm8eCDD/Lggw/uV9e6deu4/fbbeeSRR/jd737HV7/6VR577DEg2Lj+Ix/5CBs3biSdTvO9733vEP5QIiIiMl6V3DB++rjvenj5iYmpK9dLEmP+zNPYff7nqasZ+U88MER66aWXctddd3Hbbbfh7nziE5/goYceIhKJsG3bNnbs2MHcuXPL1vHQQw/xsY99DIBTTz2VU089dfDcd7/7XW699Vby+TwvvfQSmzZt2u/8UL/61a9429veRl1dHQBvf/vbefjhh7nkkktYvHgxK1asAOCMM85gy5Yt4/3LiIiIyCFQT9tEswjmBWpiUTozuVF7yS677DJ+9rOf8Yc//IG+vj5WrlzJN7/5Tdra2li3bh3r169nzpw5ZDKjb0RvZsOOvfDCC/zzP/8zP/vZz9iwYQNvfvObx6xntLbW1NQMfo5Go+Tz+VHrEhERkYmlnjaAi2+cuLq6d0LnNvKNJ1Fsz5LNF6mJR8sWra+v57zzzuODH/zg4ASEjo4OZs+eTTwe58EHH+TFF18c9Xavfe1r+eY3v8n555/Pk08+yYYNGwDo7Oykrq6OpqYmduzYwX333cd5550HQENDA11dXcyaNWtYXe9///u5/vrrcXfuuece7rzzzkP8g4iIiMhEUGibaLEkACmCtdoy+cKIoQ2CIdK3v/3tgzNJ3/3ud/PWt76VVatWsWLFCk4++eRRb3fdddfxgQ98gFNPPZUVK1Zw1llnAXDaaadx+umns2zZMo499ljOPvvswWuuueYaLr74Yo466qj9nmtbuXIl73//+wfr+NCHPsTpp5+uoVAREZEpwEYbEjtcrFq1yoeuXfbUU0+xZMmSib9ZIQc7nqTYMI8nO2qY05hkTmNy4u8zRVXs7yoiInKEMLN17r5q6HE90zbRonGIxIjkM9TEImRy2tJKREREDp1CWyXEUpDvIxmPkskVq90aEREROQwotFVCPAW5DMlYhGy+OOaWViIiIiJjOaJDW8We54snAac2ksdx+vNHxhDpkfB8pIiISLUcsaEtmUyye/fuygSNcDurGhuYQXr4D5G6O7t37yaZPHImXYiIiEymI3bJjwULFtDa2kpbW9vEV+4OHW14TYad/TX07ozRlIpP/H2mmGQyyYIFC6rdDBERkcPSERva4vE4ixcvrtwNvvxBmHEMf7PzIxzVlOT2D4y8fZSIiIjIWCo6PGpmF5nZM2a22cyuL3O+xsy+E55/xMwWhcfPMrP14etxM3tbyTVbzOyJ8NzaoXVOGXOWwY6NnDS3gWd3dFe7NSIiIjLNVSy0mVkU+DJwMbAUuMrMlg4pdjWw192PB74I3BQefxJY5e4rgIuA/zCz0l7B8919RbmF56aM9ELoeomT5tSzrb2Prkyu2i0SERGRaaySPW1nAZvd/Xl3zwJ3AZcOKXMpcEf4+W7gAjMzd+9194EdyYOpmNNNMg3FPEubgy2s1NsmIiIih6KSoW0+sLXke2t4rGyZMKR1AM0AZvYKM9sIPAFcWxLiHPipma0zs2sq2P5Dk0oDcGJTsNzHMy93VbM1IiIiMs1VciKClTk2tMdsxDLu/giwzMyWAHeY2X3ungHOdvftZjYbeMDMnnb3h4bdPAh01wAcffTRh/I7Dk4yCG1z433UJaI8u0OhTURERA5eJXvaWoGFJd8XANtHKhM+s9YE7Ckt4O5PAT3A8vD79vB9J3APwTDsMO5+q7uvcvdVLS0th/xjxi3saYv0d3Di3Ab1tImIiMghqWRoexQ4wcwWm1kCuBJYPaTMauB94efLgZ+7u4fXxADM7BjgJGCLmdWZWUN4vA64kGDSwtSTmhG8Z9o5aU4Dz+zo0o4BIiIictAqFtrCZ9A+CtwPPAV81903mtkNZnZJWOxrQLOZbQb+BhhYFuQc4HEzW0/Qm/Zhd98FzAF+ZWaPA78HfuzuP6nUbzgk4fAofe2cOKeBPT1ZdnVnq9smERERmbYquriuu68B1gw59o8lnzPAO8tcdydwZ5njzwOnTXxLKyAcHiXTzklzGwB4dkcXLQ01VWyUiIiITFdH7N6jFZdoAItA377Q9rSeaxMREZGDpNBWKZEIJJsg086s+hqa6xI8q9AmIiIiB0mhrZKSaejbC8CJ4WQEERERkYOh0FZJqTT0tQNw0twGntvRRbGoGaQiIiIyfgptlZSaAZl9oa0nW2Bbe1+VGyUiIiLTkUJbJSX39bSdOCeYjKBFdkVERORgKLRVUio92NN24px6AD3XJiIiIgdFoa2SBnra3GlIxpmfTqmnTURERA6KQlslpdLgBch2A8Fzbdo4XkRERA6GQlsllWxlBcFzbX9s6yZXKFaxUSIiIjIdKbRV0sBWVuFabYuaa8kVnJ1d/VVslIiIiExHCm2VlJoRvIeTEWY3BvuO7ujMVKtFIiIiMk0ptFXSkOHR2Q1JAHZ2qqdNRERExkehrZIGhkcHetoagp62ti71tImIiMj4KLRV0pCetub6GiKGnmkTERGRcVNoq6SaBrDoYE9bNGLMqq/R8KiIiIiMm0JbJZlBsmmwpw2CyQg7NDwqIiIi46TQVmmp9OCSHxBMRlBPm4iIiIyXQlulpWYMDo9CMBlBz7SJiIjIeCm0VdrA/qOh2Q017O7pJ69dEURERGQcFNoqLZXev6etMYk77O7JVrFRIiIiMt0otFVamZ420K4IIiIiMj4KbZWWSkOmA9yBoKcNtCuCiIiIjE9FQ5uZXWRmz5jZZjO7vsz5GjP7Tnj+ETNbFB4/y8zWh6/HzextB1rnlJNMgxegvwvY19OmyQgiIiIyHhULbWYWBb4MXAwsBa4ys6VDil0N7HX344EvAjeFx58EVrn7CuAi4D/MLHaAdU4tQ7aymlU/ENo0PCoiIiIHrpI9bWcBm939eXfPAncBlw4pcylwR/j5buACMzN373X3fHg8Cfg46pxaUjOC93CttkQsQnNdgh0aHhUREZFxqGRomw9sLfneGh4rWyYMaR1AM4CZvcLMNgJPANeG5w+kzqllyP6jAC0NNdo0XkRERMalkqHNyhzzAy3j7o+4+zLgTOAfzCx5gHUGFZtdY2ZrzWxtW1vbOJo9wYYMj0IwGUHPtImIiMh4VDK0tQILS74vALaPVMbMYkATsKe0gLs/BfQAyw+wzoHrbnX3Ve6+qqWl5RB+xiEq09M2u0GbxouIiMj4VDK0PQqcYGaLzSwBXAmsHlJmNfC+8PPlwM/d3cNrYgBmdgxwErDlAOucWsr0tM1prKGtu59isWwnoYiIiMgwsUpV7O55M/socD8QBW5z941mdgOw1t1XA18D7jSzzQQ9bFeGl58DXG9mOaAIfNjddwGUq7NSv2FCJOrBokN62pIUis7uniwt4RIgIiIiIqOpWGgDcPc1wJohx/6x5HMGeGeZ6+4E7jzQOqc0s7KbxkOw7IdCm4iIiBwI7YgwGVLpwSU/AGY3aoFdERERGR+FtskwbP/RYCurNk1GEBERkQOk0DYZUun9hkdbGrQrgoiIiIyPQttkGNLTloxHaUrFtSuCiIiIHDCFtskwpKcNwrXa1NMmIiIiB0ihbTIk05DpgGJx8NDsxhpNRBAREZEDptA2GVIzwIuQ7Ro8NLshqV0RRERE5IAptE2GVJmtrBpraOvqx127IoiIiMjYFNomw+D+oyVrtTUkyRaKtPfmqtQoERERmU4U2iZDmf1H9+2KoCFSERERGZtC22RIlhke1VptIiIiMg4KbZOhXE9bY7ArgiYjiIiIyIFQaJsMo/S07VBPm4iIiBwAhbbJkKiDSHy/nra6mhj1NTH1tImIiMgBUWibDGbBEGnf8F0R2jQRQURERA6AQttkSab3W/IDgo3jNRFBREREDoRC22Qpt/9oY1JLfoiIiMgBUWibLMnhw6NzGmrY0ZnRrggiIiIyJoW2yVK2p62GTK5IV3++So0SERGR6UKhbbKU6Wmb3aC12kREROTAKLRNltQMyHRAsTh4SLsiiIiIyIFSaJssqTTg0N85eGhgVwQt+yEiIiJjqWhoM7OLzOwZM9tsZteXOV9jZt8Jzz9iZovC428ws3Vm9kT4/rqSa34R1rk+fM2u5G+YMMlyW1mFuyJ0qqdNRERERlex0GZmUeDLwMXAUuAqM1s6pNjVwF53Px74InBTeHwX8FZ3PwV4H3DnkOve7e4rwtfOSv2GCTWw/2jJWm0NNTGS8YieaRMREZExVbKn7Sxgs7s/7+5Z4C7g0iFlLgXuCD/fDVxgZubuj7n79vD4RiBpZjUVbGvlldl/1MyY3aC12kRERGRslQxt84GtJd9bw2Nly7h7HugAmoeUeQfwmLuXJpvbw6HRT5mZTWyzKyQ1fHgUgskImoggIiIiY6lkaCsXpoauIjtqGTNbRjBk+j9Kzr87HDZ9Tfj687I3N7vGzNaa2dq2trZxNbwiyvS0AcxpTGp4VERERMZUydDWCiws+b4A2D5SGTOLAU3AnvD7AuAe4L3u/seBC9x9W/jeBXyLYBh2GHe/1d1XufuqlpaWCflBhyQ1I3gf0tMW7D+q0CYiIiKjq2RoexQ4wcwWm1kCuBJYPaTMaoKJBgCXAz93dzezNPBj4B/c/dcDhc0sZmazws9x4C3AkxX8DRMnnoJoYvgCu401dPfn6c1qVwQREREZWcVCW/iM2keB+4GngO+6+0Yzu8HMLgmLfQ1oNrPNwN8AA8uCfBQ4HvjUkKU9aoD7zWwDsB7YBny1Ur9hQpkFQ6TDnmnTrggiIiIytlglK3f3NcCaIcf+seRzBnhnmes+B3xuhGrPmMg2TqpUer8lPwCa6xMA7O7JsmhWXTVaJSIiItOAdkSYTGX2H02n4gB09uWq0SIRERGZJhTaJlNq+PBoujboaWvvy1ajRSIiIjJNKLRNpjI9bU1hT1t7r3raREREZGQKbZMpNWNYT1tjMnisUKFNRERERqPQNplSach0QrE4eCgWjdCQjNGhZ9pERERkFAptkymZBhz6O/Y73JSKK7SJiIjIqBTaJlOq/FZW6do47b2aiCAiIiIjU2ibTMmm4H3oDNJUQj1tIiIiMiqFtsk0sGl8ZvjwaLtCm4iIiIxCoW0yjTA82lQbp0OzR0VERGQUCm2TacTh0WAigrtXoVEiIiIyHSi0TaYRhkfTtXHyRacnW6hCo0RERGQ6UGibTIk6iMRG2RVBM0hFRESkPIW2yWQWDJFmhoa2YP9RzSAVERGRkSi0TbZkuuzwKKDJCCIiIjIihbbJlhpl03j1tImIiMgIRg1tZvaeks9nDzn30Uo16rBWZnh0sKdNoU1ERERGMFZP29+UfP6XIec+OMFtOTKUGx4Nn2lr1/CoiIiIjGCs0GYjfC73XQ5EmeHRZDxCIhqhvU+zR0VERKS8sUKbj/C53Hc5EAPDoyUL6ZoZTbVxOjU8KiIiIiOIjXH+ZDPbQNCrdlz4mfD7sRVt2eEqmYZiHnK9wbptoXQqruFRERERGdFYoW3JpLTiSFK6/2hJaGtSaBMREZFRjDo86u4vlr6AbmAlMCv8Piozu8jMnjGzzWZ2fZnzNWb2nfD8I2a2KDz+BjNbZ2ZPhO+vK7nmjPD4ZjO7xcym17N1I+0/WhvX7FEREREZ0VhLftxrZsvDz0cBTxLMGr3TzP56jGujwJeBi4GlwFVmtnRIsauBve5+PPBF4Kbw+C7gre5+CvA+4M6Sa74CXAOcEL4uGutHTikj7D/alEootImIiMiIxpqIsNjdnww/fwB4wN3fCryCsZf8OAvY7O7Pu3sWuAu4dEiZS4E7ws93AxeYmbn7Y+6+PTy+EUiGvXJHAY3u/lt3d+AbwGVj/cgppXR4tEQwPKrZoyIiIlLeWKGttOvnAmANgLt3AcUxrp0PbC353hoeK1vG3fNAB9A8pMw7gMfcvT8s3zpGnVPbKMOjPdkCucJYf1YRERE5Eo01EWGrmf0lQThaCfwEwMxSQHyMa8s9azZ0mZBRy5jZMoIh0wvHUefAtdcQDKNy9NFHj9HUSTTC8Gjprgiz6msmu1UiIiIyxY3V03Y1sAx4P3CFuw90D70SuH2Ma1uBhSXfFwDbRypjZjGgCdgTfl8A3AO8193/WFJ+wRh1AuDut7r7Kndf1dLSMkZTJ9FAT9tI+49qBqmIiIiUMWpPm7vvBK4tc/xB4MEx6n4UOMHMFgPbgCuBdw0ps5pgosFvgcuBn7u7m1ka+DHwD+7+65L7vmRmXWb2SuAR4L0M315raotEoWb4/qMDoa1DuyKIiIhIGaOGNjNbPdp5d79klHP5cFP5+4EocJu7bzSzG4C17r4a+BrBTNTNBD1sV4aXfxQ4HviUmX0qPHZhGCKvA74OpID7wtf0kmwqMzwa7D+qGaQiIiJSzljPtL2KYKLAtwl6tsa1Jpq7ryGcvFBy7B9LPmeAd5a57nPA50aocy2wfDztmHJSTcOGR9MaHhUREZFRjBXa5gJvAK4iGNr8MfBtd99Y6YYd1pLpEYdHFdpERESknLF2RCi4+0/c/X0Ekw82A78IZ5TKwSozPNqY2jd7VERERGSosXraMLMa4M0EvW2LgFuA71e2WYe5VHrY8Gg0YjQmYwptIiIiUtZYExHuIHh+7D7gsyW7I8ihKDM8CtBUq10RREREpLyxetr+HOgBTgQ+VrI3uwHu7o0VbNvhK5mGXC/ksxBLDB5Oa/9RERERGcFY67SNtfiuHIxUya4I9fsW/k3XxmlXaBMREZEyFMqqYXArq/2HSBtTcTo0e1RERETKUGirhhG2skqn4hoeFRERkbIU2qohNfKm8e19Ody9Co0SERGRqUyhrRpGGB5tSsUpFJ3u/nwVGiUiIiJTmUJbNQwOj+7d73A6pf1HRUREpDyFtmoYYXi0qVZbWYmIiEh5Cm3VEKuBWGrY8GhaW1mJiIjICBTaqiXZNGz26EBPm0KbiIiIDKXQVi2p9PDZo+EzbRoeFRERkaEU2qqlzP6j6YFn2vq0/6iIiIjsT6GtWsoMjybjURKxiHZFEBERkWEU2qqlzPAoaFcEERERKU+hrVrKDI9CuCuCetpERERkCIW2akmlIdMJxeJ+h5tScT3TJiIiIsMotFVLsglw6O/c73BTKkFHn7axEvPB15AAACAASURBVBERkf0ptFXLCPuPpmvjdPSqp01ERET2V9HQZmYXmdkzZrbZzK4vc77GzL4Tnn/EzBaFx5vN7EEz6zazfx1yzS/COteHr9mV/A0VM7CV1dAFdlNx2jURQURERIaoWGgzsyjwZeBiYClwlZktHVLsamCvux8PfBG4KTyeAT4FfHyE6t/t7ivC186Jb/0kGNg0ftgCu3F6swWy+WKZi0RERORIVcmetrOAze7+vLtngbuAS4eUuRS4I/x8N3CBmZm797j7rwjC2+FplOFR0FZWIiIisr9Khrb5wNaS763hsbJl3D0PdADNB1D37eHQ6KfMzCaisZNuhOHRxsFN4/Vcm4iIiOxTydBWLkz5QZQZ6t3ufgrwmvD152VvbnaNma01s7VtbW1jNnbSjTQ8WhvsP6qeNhERESlVydDWCiws+b4A2D5SGTOLAU3AntEqdfdt4XsX8C2CYdhy5W5191XuvqqlpeWgfkBFJerBosOHR8OeNi2wKyIiIqUqGdoeBU4ws8VmlgCuBFYPKbMaeF/4+XLg5+4+Yk+bmcXMbFb4OQ68BXhywls+GcyCIdK+ETaNV2gTERGRErFKVezueTP7KHA/EAVuc/eNZnYDsNbdVwNfA+40s80EPWxXDlxvZluARiBhZpcBFwIvAveHgS0K/Dfw1Ur9hopLNg0bHm1KaSKCiIiIDFex0Abg7muANUOO/WPJ5wzwzhGuXTRCtWdMVPuqrsz+ow3JOGZorTYRERHZj3ZEqKYyw6PRiNGY1K4IIiIisj+FtmoqMzwKwRCphkdFRESklEJbNZUZHoVgMoKGR0VERKSUQls1DQyPDpkw25SKa/aoiIiI7EehrZqSTVDMQa53v8NNqTid6mkTERGREgpt1TS4/+jQXRE0PCoiIiL7U2irphH2H02nErT3ZikWx9rRS0RERI4UCm3VNLj/6P6hrSkVp+jQnc1XoVEiIiIyFSm0VdMIw6NN4VZWHZqMICIiIiGFtmoacXhUW1mJiIjI/hTaqmmwp2348Cho03gRERHZR6Gtmgafadt/eHRuUxKAP+3pHXqFiIiIHKEU2qopEoWaxmHDo0fPrKUpFWdD6/DdEkREROTIpNBWbWW2sjIzTl3QxOOtw/clFRERkSOTQlu1jbBp/IqFaZ7d0UWvlv0QERERFNqqb2D/0SFOW5CmUHQ2bu+sQqNERERkqlFoq7Zk07DhUYBTFwaTFB7fqufaRERERKGt+pLpssOjsxuSzE+nWK/QJiIiIii0Vd8Iw6MApy1s4nHNIBUREREU2qovmYZcDxSGL6R72oI0W/f0sbu7vwoNExERkalEoa3aRlhgF+C0hcGOCRu2aekPERGRI51CW7WNsP8owCnzm4iYJiOIiIiIQlv1Nc4L3tu3DDtVVxPj+Nn1Cm0iIiJS2dBmZheZ2TNmttnMri9zvsbMvhOef8TMFoXHm83sQTPrNrN/HXLNGWb2RHjNLWZmlfwNFTdnWfD+8pNlT5+2IM3jrR24+yQ2SkRERKaaioU2M4sCXwYuBpYCV5nZ0iHFrgb2uvvxwBeBm8LjGeBTwMfLVP0V4BrghPB10cS3fhKlZkDTQtgxQmhbmGZPT5bWvX2T3DARERGZSirZ03YWsNndn3f3LHAXcOmQMpcCd4Sf7wYuMDNz9x53/xVBeBtkZkcBje7+Ww+6nr4BXFbB3zA55iyHl58oe2pFOBlB67WJiIgc2SoZ2uYDW0u+t4bHypZx9zzQATSPUWfrGHVOP3OXw67nIJcZduqkuQ0kYhE2aL02ERGRI1olQ1u5Z82GPph1IGUOqryZXWNma81sbVtb2yhVTgFzloMXoO2pYafi0QjL5zXy+FYt+yEiInIkq2RoawUWlnxfAGwfqYyZxYAmYM8YdS4Yo04A3P1Wd1/l7qtaWlrG2fRJNveU4H2kyQgL0zyxrYN8oTiJjRIREZGppJKh7VHgBDNbbGYJ4Epg9ZAyq4H3hZ8vB37uo0yTdPeXgC4ze2U4a/S9wA8nvumTbMZiiNeN+lxbX67Aczu7J7lhIiIiMlXEKlWxu+fN7KPA/UAUuM3dN5rZDcBad18NfA2408w2E/SwXTlwvZltARqBhJldBlzo7puA64CvAyngvvA1vUUiwdIfI80gXRBMRnh8aztLjmqczJaJiIjIFFGx0Abg7muANUOO/WPJ5wzwzhGuXTTC8bXA8olr5RQxdzk88T1whyFLzx3TXEtTKs7jre1cedbRVWqgiIiIVJN2RJgq5iyH/g7o2DrslJlx6oIm1msygoiIyBFLoW2qGGMywoqFaZ7d0UVftjCJjRIREZGpQqFtqpi9FLARJyOctiBNoehs3K7eNhERkSORQttUUVMPMxfDjhFC28I0ZnD3utay50VEROTwptA2lcw9ZcTh0ZaGGq55zbHc9ehWfvLkS5PcMBEREak2hbapZM4psPcF6O8qe/pvLzyJUxc08fffe4Lt7dpAXkRE5Eii0DaVzA1XMtmxsezpRCzCLVeeTr5Q5K+/s55CcbQdv0RERORwotA2lcwJQ9sIkxEAFs2q458uW87vX9jDlx/cPEkNExERkWpTaJtKmhZAsmnEnREGvH3lAi5bMY+b//tZ1m4ZbatWEREROVwotE0lZjD31BEnI5T6p8uWs2BGLX9113o6+nKT0DgRERGpJoW2qWbOcti5CYqjL6LbkIxzy1Wns6Mzw0U3P8Rtv3qB3mx+khopIiIik02hbaqZuxxyvbDn+TGLrliY5s6rX8HCGbXccO8mzrnpQf7158+p501EROQwpNA21RzAZIRSrzqume9e+yr+77Wv4rQFTfzzT5/l7Bt/zi0/e47+vLa8EhEROVwotE01LSeDRcecjDDUmYtmcvsHzuLevzyHc46fxRceeJa33PIr1r24t0INFRERkcmk0DbVxJMw68QDmoxQzvL5Tfz7n5/Bbe9fRU9/nsv//Td8+odP0t2v591ERESmM4W2qWjuKQc8PDqS1508h5/+zbm871WL+MbvXuQNX/glP3tqxwQ1UERERCabQttUNHc5dG2H3kNbg62+JsZnLlnG9657NQ3JGFffsZZr71zHSx3aAktERGS6UWibigYmIzyzZkKqW3n0DO79y9fwd288iQef2cnr/79fctuvXiBfKE5I/SIiIlJ5Cm1T0TFnw/xVsPpjsP7bE1JlIhbhI+cfzwP/81xWLZrJDfdu4rJ/+zWP/UkTFURERKYDhbapKJ6E9/4QFp0DP7gWfvfvE1b10c21fP0DZ/Lld61kZ2c/b/u33/Du//wdv3hmJ+7agF5ERGSqsiPhX9SrVq3ytWvXVrsZ45fvh7s/CE/fC+f9A5z798FWVxOkK5PjW4/8idt/vYWXOzOcNKeBD71mMZesmEdNLDph9xEREZEDZ2br3H3VsOMKbVNcIQ8/+his/ya84lp44/8LkYntIM3mi9y7YTu3PvQ8T7/cxaz6Gt6+cj7vPGMBJ8xpmNB7iYiIyOiqEtrM7CLgS0AU+E93v3HI+RrgG8AZwG7gCnffEp77B+BqoAB8zN3vD49vAbrC4/lyP2qoaR3aAIpF+Okn4Xf/Bse9Di75V2iaP+G3cXd+tXkX3/jtizz49E7yRee0hWkuP2MBl5w6j6ba+ITfU0RERPY36aHNzKLAs8AbgFbgUeAqd99UUubDwKnufq2ZXQm8zd2vMLOlwLeBs4B5wH8DJ7p7IQxtq9x914G2ZdqHNgB3WHsb/PR/QSQOF98Ep105ocOlpXZ19/ODx7Zx97pWnn65i2Q8wscvPIkPnL2YaKQy9xQREZGRQ1slJyKcBWx29+fdPQvcBVw6pMylwB3h57uBC8zMwuN3uXu/u78AbA7rO3KZwZlXw3W/hjlLgwkKd70bundW5Haz6mv40GuO5b6/es3g1lif+/FTvPPff8Pmnd0VuaeIiIiMrJKhbT6wteR7a3isbBl3zwMdQPMY1zrwUzNbZ2bXVKDdU9vMY+H9P4YLPw+b/xu+/Ar47Zehu60itzMzls9v4qvvXcWXrlzB87t6eNMtD/OVX/xR67yJiIhMokqGtnJjaEPHYkcqM9q1Z7v7SuBi4CNm9tqyNze7xszWmtnatrbKBJqqiUTh1R+Fax+GlpPg/k/AF06Gb18Fm1ZDPjvhtzQzLl0xnwf+57m87qTZ3PSTp3nbv/2GO36zhU3bOykUD/8JLSIiItUUq2DdrcDCku8LgO0jlGk1sxjQBOwZ7Vp3H3jfaWb3EAybPjT05u5+K3ArBM+0TcDvmXpaToIP/gR2PgXrvwUbvhPsopCaCadeAas+CC0nTuwtG2r4yntWsuaJl/nfa57i06s3AtCQjHHGMTM4c9FMLlo+l+Na6if0viIiIke6Sk5EiBFMRLgA2EYwEeFd7r6xpMxHgFNKJiK83d3/zMyWAd9i30SEnwEnAEkg4u5dZlYHPADc4O4/Ga0th8VEhANRyMPzD8Jj/wVP/xiKOVj0GjjzQ3DymyFaMvuzWITe3RCNQWrGQd3O3dnW3sejW/bw6Ja9rN2yh2d3BM+7rTpmBn925kLefMpR1NVU8r8NREREDi/VWvLjTcDNBEt+3ObunzezG4C17r7azJLAncDpBD1sV7r78+G1nwQ+COSBv3b3+8zsWOCesPoY8C13//xY7ThiQlup7p3wh2/Auq9Dx1aonwsLVkH3Duh8CbpfhmI+mIl62hXw6o8FPXeHaGdnhnse28Z31m7l+bYe6hJR3nLqPC49fR5nLZpJLKpNOEREREajxXWPtNA2oFiA5x6AdbfD3hehYS40HAUNc4L3Xc/CY9+EfB+c9CY4+6/g6Fce8m3dnXUv7uW7a7dy74aX6M0WaErFueDk2Vy4bA6vPbGF2oR64ERERIZSaDtSQ9uB6NkFv/8q/P5W6NsDc06B5uPCcDcXGucFn2csCj5HxrfFVW82z8PP7eKnG3fws6d30N6boyYW4aLlc7nuvOM4eW5jZX6XiIjINKTQptA2tmzPvufhOrdD18uQ7dq/TDQB6WNg5mKYsRjSC6EpfKUXQl3LqAv+5gtFHt2yl588+RJ3r2ulJ1vg9Uvm8JHzj+P0ow/u2ToREZHDiUKbQtvB6e8KwltHK+x9Afa8EL5vgb1bhoe6WBIWnwunvBNOfhMk6vY/3/4nePJ78NSPyKVauLv+3dy0IUl7b46zj2/mnWcs5JjmWhbMqGVWfQKr0I4PIiIiU5VCm0LbxHOHTHsQ6Nq3Bu+7N8PT90LnNojXBrNWl70tOP/k96D198G1804PQl/fXvInXMTqGe/nxsfi7OzqH6w+FY+yYEaK0xam+asLTmDhzNrq/E4REZFJpNCm0DZ5ikX402/hif8Lm34AfXuD43OWw/J3wPK3B8/HZTrhkf+A3/4LZDoonvxWti75C56LnUhre4ate/v4055eHn6uDXe45rXHct15x2kCg4iIHNYU2hTaqiOfhRd/HUxkmH1y+TJ97fC7r8Dv/g36O4PlSU58YzCb9dhz2d4DN973NKsf387cxiTXX3wyl66YNzh06u5kckUK7tRrTTgREZnmFNoU2qa+vnZ49ifBrg6bfx48LxdLwcIzoelotvkMvr/ZWbc3Ra5xEdsi8+jqL9CZyZErOBGDt69cwF+//gQWzNBQqoiITE8KbQpt00u+H7b8Kghx29btWxDY921SvzfWwh8bX0Fr86vZPfvVbO1L8K3f/wkc3nvmHD58VpqZ8TzMOmHUGa0iIiJTiUKbQtv0V8iHOzpsh50b4Y8/hz/+Avo7wCIweyn5TDf5rjaSxZ7By15uOIX75n2EDZEl7O7J0tGXY+GMFKctSHPqgiaWz2/SVlsiIjJlKLQptB2eCnnYthY2/wy2PwbJJqhrYQ+N3LelwHOtO7k29iPm2l5+GXkF32n8IJ31i3lhVw/b2vsAiBgc31LHrPoaYrEosYgFr6iRrk0wP51ifjrFvHSKeekksxuSJGLajktERCpDoU2h7YjU0ZejpthHcu1/wK9vhlwfrHwvNBxFX9sL9LW9QKSzlfrMy8TIkydKnhgFIhSIsZYlfKb/Kv7kc/art6Emxoy6BDNq48yoS3BUU5KT5zay5KhGTj6qgcZkvEq/WEREpjuFNoU26d4Jv7gR1n0dvBDMUk0vhPTR0LQgWBi4kINiPnhle+DJ7+GFLO2nXcOm4/+Crd0R2rr62dObZW9Plj29Odp7s2zd08ve3tzgrRbMSHH87HoaknHqa6LU18Soq4lRm4gSMcPMMIJevmjEaKpNMKsuwcz6BM11NcyojROLHlxvXrHoFNyJRUyLE4uITEMKbQptMiDTAdEaiCfHLtv1Mvz3Z+DxbwfLlrz+s7D0Etj1HOx8CnZugran8f5O+hoWsz22kKfzc1nbPYs/dDbQlYWuTJ6e/jx9ucK4mhmNGBEDs+A9YkY8GqEmFiEZj3K8bef6zBfZY2luil/H1lwjPf2FwfuYQSIaCV6xCE2pOCfOaeDEuQ2cPLeBk+Y2cFRTkraufl7uyPByZ4YdnRl2dWfpyxbI5IK6MrkiuUKReDRCKhElGQveU/EoLQ01zG1KclRTkrlNKeY01Bx02BQRkYBCm0KbHIqtj8J9fxc8N1cqEg9mp9Y0BLtB9O7e/3wsBfEUxGvxeJJirBZPzaSYmonXNlNMNZNLzqTXU7QXkuwtJNidS9CWS7I7Po8iUPRgLbpC0ckVivTni5yy+z7e8dIXyFmCGu8nG0nyw4X/Dy/OvoDaRJSoWVC2UCSXd7KFAru6sjyzo4stu3sY7f/sa2IRasNQloxHqYlHScQiZPNF+geDXIGebIFsvrjftRGDGbUJZtQlmFmXYGZt0Hs4ozbOjNoE6drgc7o2QcSgvS9HR9hb2d6XI5MrEo8a0UgQUGMRoyYWoTEVpzEZp6k2eG9IxkjEgjA6EEwjkfH1Kro7Xf152rr6aevqJxYxmutraK5P0FATUy+liFSNQptCmxyqYjHY5WHPH6HlZJi9FJqPg2jJ82u9e4JeuF3PQsdWyPUGz9HlMsHnbHdQpndX8N7fOfL9ZiyG098DK94NjUcFx7K9sObvYP1/wTFnwzv+E/q74fsfgpceD8pfdGMQIktlOoPlUlJp+rIFNu/s5umXO9nZ1c/ssLdsbmOSOU1JGujFMp3BvrP9XcF6eflssPXYQDsIQk9nJs/LHRm2d/TxckeGlzoy7O7uZ09Plt094RByT5a9vVmKB/D/ahLRCLlicdRQOZJoZN9iyw6DdSRiEVLxIITWJoIQ2pXJ0dbVT/+Q0Fnajpl1ieHhMBahIRljTuO+3sW5jUnqa2Ls7c2yq7uf3d3Bb+7uzzOrPsHshiSzG2uY3ZBkZl2C9r4sOzv72dnVz86uDLu7s8SjEeprotQmYoND6Q3JgVecxvDdDHKFItl8kVzByeaLZAvFkmPBu5nRkAzqqg/fa2IRBVGRaUKhTaFNpqJ8fxDest0lIak7WNrkye/DloeD5UxOuBCWXAK/+Rdoexpe+3E493qIhkuV5LPwyxvh4S/AjGOCyRZ7Xwx6/3ZvDuqzKCw6J9gLdslboW5WcK17MMz71L3w9I/g5SdGbm/LEjjudXDc+XDMqyFRd0A/s1gMerXae7Ps7c2xtzeLu5OuTZBOBT1vjcnY4NBqsejkikXyBSeTK9CVydPRl6Mzk6OjL0dXJj8YUPrDsJIrBAHMCIaTB9bmy+aL9GWD4em+XJG+bIH6miizG5O01NfQ0lDDrPoaCu7s6QlC167uLLu7++nuzw8Go/7wXl2ZHC93ZOjNlg53O7AvEEUjRm08Sld/fsy/TW0iSr7gZAvlA+REiUWMZDxKMh4J3/cF2fqaGLU1MeprotQlYsEweDw6WD4V9rbGS4bbYxEjXwz+98nkisF7vkB/Lvh7ZfPFwb9doUxij0WNukQQUOsSUerCYDnYsxz2LkcjxszaBLMaamiuSzCjNjHuXlWR6UahTaFNpqPdf4TH7oT13wqCV10LvP3WIDiV8+Jv4Pv/Azr+BLXN0Hw8NJ8Q9Ahmu2HjD4KewoEAN3sJPHs/7H0BMFh4FpzwBqifE/TWJRqCd7Og7ucfhBd/C4V+iCZg0Wvg5DcFW441ztu/LR2t8Pwv4IWHAYeWk2DWSUEv5YxFQZ17twTPBrY9HbyK+aBdi88N2l6tniF32PYH2PFk0Cvaszt83wWZDjzbjfcHL8v1UIim2LP4zWRPuYr6xa+kKQwWmVyBtrBHbUdn0AM5ozYR9rwFgXFgL91svkhPf57u/jw92TxdmTxdmSCgdoaf3YPh63gYnOLRCPGoDTtWdKenv0BXJkd3f1BXd3+e/lyRTD4Y3u7PFenLFYL7DbyyhcHnLyfiXw0RIwx4w59zHAh24xWNGOlUnJpYhFg0QixqxCPBeyxiRCJG1IL3gSHvYNmeJPNnpJifriUWtfA3F+jNBn+bfGHfDx74xy5XKA7+7boyebozwd9m4FlTK3nWdHZDDXMak8xtCt5nNyTDyUD7ygb/LbHvn+mBTwV3CgUnX3Ty4X+spBJRmlJx4mWeES0Wnfa+3GDP7sDjBXt7s7T35igWnWNb6jlhTj3Ht9Qzoy4x7r+zVJdCm0KbTGeFPGx9JAg+Az1kI5bNBT12tTOHn3MPgsjGH8DGe6D9T7D4tbDkLXDSm6FhzvBrhsr1BQHujz8Pthzb83xwfN7pcOJF0NMWhLXdm4PjdS3BxI/O1n11RBOABeFvQNPCYAi3c1vwveGooG3zVkIsEfQ4DrwGrs33Qz4TvvcH1w++PJgl3N8dLMDc3xUME2e7YeZxQThcdA7MPQUi0eCeOzbCE3fDk9+D9hf3tS1RH/w9a2dBKh18T9RDTX3Q29ixDZ76EeT7gmC64l1w6p8ND7IjcQ/aXMxDsRC0O1FftdDqHvT8ZbJFsu3bsWfvI9OwiI6WlWSJDw7PxqI22GM30INXUzKcPNaklFyhSG9/ge5sEBr7c0UsnFE9MBGnUITdYQ/o7u5+dodD77mBHtaiky8EQSdfdIphD12hGHxv6+rnpY4+coVD+3ddbSJKQzJGMh4N/ufCKYaZsz9fYHdPdkKC7lD1NTGaUnFm1MXJF5zd4SMH5XovIQj1QZv2BeJZ9QmOaa4jGd/XWxoveR40HrP9elGTYQ9sbSJKKhH0hDYk4zSl4qRrg/dkPLrffd2Dv7cR/O9XGk7zhSLb2vt4YVcPL+7uZcvuHrozeY5prmXxrHoWzapl8ay6wf+AEYU2hTaRodyDkBA9hDXl3KHtGXjmx/D0mmCh43hdEIaOPS94zV4ShI9MZ/i83zNBr5oXg+HWlpOh5cSgR889CIEvPLTv1bvrwNoSiQU9iKXhziJBqEo2Qk1j8B5PwY5NQY8jBAsyH/2qYDi57amgjmPPg1MuD3oS62YF14wl0xkE4fXfDAI2QDINMxcHzyfOXBwE0Z5dwfOOHVuhfWuww0dpeB1QPzcYgl50NhxzThDYS0NcLhPMhC7mgt8eiQfD5ZF4sHxNmd4tIBhK3/EEtK4Lgum80+HoVwbL3gwoFmHLQ/Do14JgXgyHeeN1cOy5cPwFcNwF+3pMp7hCGN62tfeyrT2Du1ObiFEXDgfX1cRIhAEzeCIyEI1YuGxPbPCZyZHkCkV2hjOxg1nY/RSKHga88FnLkn/dDtzHPbhPLGJEoxHiYW9hJlegvTcXvoLnQqMRY1Y4WSZ4D4aM0+FEnxm1CVKJKMWis629j81t3Wze0c3mnd38aU/vfs8/DvR05gu+/7HCgT1TmogFbc0V94XkUgMLlMcjEfpyBfIl52vD4fC2rv3/uZ9ZlyAS/vM08I9VxKA2ESMVj1IXPvdZm4hiFvxj6vjgkHomV6Qnm9+vF7UmFt3396kLHsVIRCODPZq5QtC76c6wXttkPBpMgkoFQXXgtfSoxoovsK7QptAmUnm9e4IeotgEDce4Bz13pT1oxQLgQTCJ1QTv0ZqRQ8pIOrcH+9tueTjoOaxrgeXvgKWXQf3/3969x8h13mUc/z6evXh3fdn4QhNfYruqKQSUJo2xQkAoCq2U0gqDKLJDEVEIqiiFBFQuLfzBRSBRCUEbpaoUkkCDqqRVKGDxR0tII25t3bh1CU4Dres4juNNYsdeX9Z73x9//N7pTNe7rtPd2ZnxPB/paM45e+bMO/P6XT/7nvc9s35h5T55KL8399ThvPR86vns1YxpQBneVm/K+wSu2gjd/dnbt6ySoRFybOEL/wXnhnK7f1329o2dgdHhuYNelSqw8uqyXJM9flqW3+M79Eztucu6aoFs9eYMb1dtzfGUp74FfVflRJgbfiHL/80n4NATuQ4ZEPvX5HF9a3J9zbbsHd1w43eGupmZnKDz0n44tj8vn19UbmUvbFdv1mm1fnsGculdWR5X5ee3ZluG7mYZP5e3BZoYyT9Ounob/5ozMzk5aOxsTmTq7l/U8BwRjE/l2M8Lk9OMTkyVS+1TDI/m1wAOX5jk7Ogk0zORl6hLz2hXCbbV3s+c8R709Sxjy9oBtq0bYMvaftav6EUSFyamOHLyAs+fHOHIayMcHx79jklE+XaD0ckMYNUgVh1PWr08DfnYV3oHB3q66O/N9fHJGU7XBd/hC5NMTs98u9zd5RK7gMnpag9t9iSPTU7POVnp6T94G+tXNrauHdoc2sysmaanMoD2r738UFvteXzhC7lMjZYew9W1pdKTvW3T5abQM+Xy+NmhDHznhnJ9egI23AAbb4JNO2Djjgx0rxyEo1+Co1/Mx/Mvw+abYccvw3W7Lr6fYUSOtTz8VAav0VMwehounM5b3pz6Vr4WZJjbcGMG7eMHarOle1dn4NKsoB0z+dyp8drj1Fhe0o55xr/1XZWhZXBLhrruvhL2+vJznhjJcl04VR5fy1DbN1jCZlkiYGw438toeZwer/1xUA2RMZNB7dxQlquq0pvvdfPODL9XX59DCUZew/x8mgAACb9JREFUzXofOZmvPTNdF7CU65MX8jJ+/YSk+sv+U2P5eUyM5M+Iiz+DjTeVZUcZwzpSznc+Q97ESPlMJ/JxeiLL0ruy9lksH8ze6ImR8jmcrn0efYMZ7Ac3w+prM6C3QS/rQo1NTnN2NCdAVQPrrW9e3/D7UTq0ObSZWSeL+O7/yUZksFpI79XURM5GPn6gtgBs+pFaWFz7ptfXMxqR4WWiBJvR4eztO32kLKUnc2IkA87kWAZcKL2Ba3MZWJs9gjFdglk1pJ0CVAJcNcCsLt+SUj92soTRag/myquzF7PSnb2YR/fB0NdqofVyVXrL+MgVpTdxRQmfyzN4di3PcF7tZay/3D86nK/90lfz8v584XaxdQ9kcOvuh57+Uub+fP3Zs+FjJgNh32D5Y2MwA/DocAnK5XFyNN//8nJc32Bt2MTMVNbbzHTtG2uq5x8/n9s9AzmcYWBd9pz3l6ENWlZ6sZfVerKJ2lhSIvd31+6rSXdf9kRPjee/pcmx8m9gDG75jcsbMrEATQltkm4HPgpUgAcj4s9n/bwXeAS4CXgN2B0RR8rPPgTcDUwD90TE5y7nnHNxaDMz6zAROSmn0r20PUKTYxncXjmYPYor1meAGFifgbFSN9g+SnB4vZf25zN+Pu/XOHy0BLwVZQZ4mTDTtTw/j0pvhkAty5A+Vhdgx89mAKvvhexdlccMH62NxTzzYj5nciTvHzkxkusog1bvqloQlWqX9auvNT1RC2bLS1Du7isThqrHnsnyfDt0VWrDCHrqQm7vigyRE+ezN3PkRFley8BVHVqxWD7wjcubtLUASx7aJFWAbwBvB44BTwN3RMTX6475NeD6iPhVSXuAn42I3ZKuAx4FdgIbgH8Fvr887ZLnnItDm5mZWQerBuSo3l9RpeetXKKenio9aqPlpuhjOdSga3ku9ZfdK42f5TpfaGvkK+8EDkXE4VKAx4BdQH3A2gX8UVl/HLhfOU94F/BYRIwDz0s6VM7HZZzTzMzMrKYazpinV7PSBZWVF3+bTItp5Ei6jcCLddvHyr45j4mIKeAMsPYSz72cc5qZmZldcRoZ2uYaRDD7Wux8x7ze/Re/uPReSfsl7T9x4sQlC2pmZmbW6hoZ2o4Bm+u2NwHH5ztGUhewGjh1iedezjkBiIgHImJHROxYv36B91wyMzMza7JGhrange2StknqAfYAe2cdsxe4s6y/G/h85MyIvcAeSb2StgHbgS9f5jnNzMzMrjgNm4gQEVOSfh34HHl7jocj4llJfwLsj4i9wEPA35WJBqfIEEY57tPkBIMp4P0ROeVjrnM26j2YmZmZtQrfXNfMzMyshcx3y4/Gfg+DmZmZmS0KhzYzMzOzNuDQZmZmZtYGHNrMzMzM2kBHTESQdAJ4ocEvsw442eDXsO+N66Y1uV5al+umNbleWlMj6mVLRFx0k9mOCG1LQdL+uWZ6WPO5blqT66V1uW5ak+ulNS1lvfjyqJmZmVkbcGgzMzMzawMObYvngWYXwOblumlNrpfW5bppTa6X1rRk9eIxbWZmZmZtwD1tZmZmZm3AoW0RSLpd0v9JOiTpg80uT6eStFnSU5Kek/SspHvL/jWSnpD0zfJ4VbPL2okkVSQdkPTPZXubpH2lXj4lqafZZexEkgYlPS7pf0vb+VG3meaT9Fvl99hBSY9KWu420xySHpb0qqSDdfvmbCNK95U88Iykty5mWRzaFkhSBfgY8A7gOuAOSdc1t1Qdawr4QET8IHAz8P5SFx8EnoyI7cCTZduW3r3Ac3XbHwb+qtTLaeDuppTKPgp8NiJ+AHgLWUduM00kaSNwD7AjIn4YqAB7cJtplr8Fbp+1b7428g5ge1neC3x8MQvi0LZwO4FDEXE4IiaAx4BdTS5TR4qIoYj4alk/R/7ns5Gsj0+Uwz4B/ExzSti5JG0C3gk8WLYF3AY8Xg5xvTSBpFXATwAPAUTEREQM4zbTCrqAPkldQD8whNtMU0TEvwOnZu2er43sAh6J9CVgUNI1i1UWh7aF2wi8WLd9rOyzJpK0FbgR2Ae8ISKGIIMd8H3NK1nH+gjwu8BM2V4LDEfEVNl2u2mONwIngL8pl64flDSA20xTRcRLwF8AR8mwdgb4Cm4zrWS+NtLQTODQtnCaY5+n5DaRpBXA3wO/GRFnm12eTifpXcCrEfGV+t1zHOp2s/S6gLcCH4+IG4ERfCm06cr4qF3ANmADMEBedpvNbab1NPR3m0Pbwh0DNtdtbwKON6ksHU9SNxnYPhkRnym7X6l2T5fHV5tVvg71Y8BPSzpCDh+4jex5GyyXfsDtplmOAcciYl/ZfpwMcW4zzfU24PmIOBERk8BngFtwm2kl87WRhmYCh7aFexrYXmb19JCDRfc2uUwdqYyTegh4LiL+su5He4E7y/qdwD8tddk6WUR8KCI2RcRWsn18PiLeAzwFvLsc5nppgoh4GXhR0pvLrp8Evo7bTLMdBW6W1F9+r1XrxW2mdczXRvYCv1Rmkd4MnKleRl0MvrnuIpD0U2TPQQV4OCL+rMlF6kiSfhz4D+B/qI2d+n1yXNungWvJX4Y/HxGzB5XaEpB0K/DbEfEuSW8ke97WAAeAX4yI8WaWrxNJuoGcINIDHAbuIv+gd5tpIkl/DOwmZ8UfAH6FHBvlNrPEJD0K3AqsA14B/hD4R+ZoIyVk30/ONr0A3BUR+xetLA5tZmZmZq3Pl0fNzMzM2oBDm5mZmVkbcGgzMzMzawMObWZmZmZtwKHNzMzMrA04tJlZx5M0LelrdcuifSuApK2SDi7W+cysc3V990PMzK54oxFxQ7MLYWZ2Ke5pMzObh6Qjkj4s6ctleVPZv0XSk5KeKY/Xlv1vkPQPkv67LLeUU1Uk/bWkZyX9i6S+pr0pM2tbDm1mZtA36/Lo7rqfnY2IneRdzj9S9t0PPBIR1wOfBO4r++8D/i0i3kJ+h+ezZf924GMR8UPAMPBzDX4/ZnYF8jcimFnHk3Q+IlbMsf8IcFtEHJbUDbwcEWslnQSuiYjJsn8oItZJOgFsqv9qIUlbgSciYnvZ/j2gOyL+tPHvzMyuJO5pMzO7tJhnfb5j5lL//ZDTeDyxmX0PHNrMzC5td93jF8v6F4A9Zf09wH+W9SeB9wFIqkhatVSFNLMrn//aMzMrY9rqtj8bEdXbfvRK2kf+kXtH2XcP8LCk3wFOAHeV/fcCD0i6m+xRex8w1PDSm1lH8Jg2M7N5lDFtOyLiZLPLYmbmy6NmZmZmbcA9bWZmZmZtwD1tZmZmZm3Aoc3MzMysDTi0mZmZmbUBhzYzMzOzNuDQZmZmZtYGHNrMzMzM2sD/A2nieSdJF3etAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the training data\n",
    "from matplotlib import pyplot as plt\n",
    "#plt.figure()\n",
    "plt.figure(figsize=(10,5))\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('LSTM Model loss')\n",
    "plt.ylabel('MSE')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Validation'], loc='upper left')\n",
    "#plt.show()\n",
    "#plt.savefig(os.path.join(OUTPUT_PATH, 'train_vis_BS_'+str(BATCH_SIZE)+\"_\"+time.ctime()+'.png'))\n",
    "plt.savefig(os.path.join(OUTPUT_PATH, 'train_vis_BS_'+str(BATCH_SIZE)+\"_\"+time.strftime(\"%a %b %d %H %M %S %Y\", time.localtime())+'.png'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<keras.engine.sequential.Sequential object at 0x000000000529B808>\n",
      "Mean Square Error is 0.0005865085250056956 (280,) (280,)\n",
      "[0.39430794 0.39276293 0.39186856 0.39317703 0.39709783 0.3936996\n",
      " 0.38725063 0.38186577 0.37664855 0.36832738 0.36099705 0.3561939\n",
      " 0.353419   0.3528223  0.3550381 ]\n",
      "[0.3877551  0.39591837 0.41632653 0.44081633 0.36326531 0.37142857\n",
      " 0.3755102  0.37142857 0.36326531 0.35102041 0.3755102  0.36734694\n",
      " 0.3877551  0.4        0.40816327]\n",
      "[1.2560545 1.2522691 1.250078  1.2532837 1.2628897 1.254564  1.238764\n",
      " 1.2255712 1.2127889 1.1924021 1.1744428 1.162675  1.1558765 1.1544147\n",
      " 1.1598433]\n",
      "[1.24 1.26 1.31 1.37 1.18 1.2  1.21 1.2  1.18 1.15 1.21 1.19 1.24 1.27\n",
      " 1.29]\n"
     ]
    }
   ],
   "source": [
    "# load the saved best model from above\n",
    "saved_model = load_model(os.path.join(OUTPUT_PATH, 'best_model.h5')) # , \"lstm_best_7-3-19_12AM\",\n",
    "print(saved_model)\n",
    "\n",
    "y_pred = saved_model.predict(trim_dataset(x_test_t, BATCH_SIZE), batch_size=BATCH_SIZE)\n",
    "y_pred = y_pred.flatten()\n",
    "y_test_t = trim_dataset(y_test_t, BATCH_SIZE)\n",
    "error = mean_squared_error(y_test_t, y_pred)\n",
    "print(\"Mean Square Error is\", error, y_pred.shape, y_test_t.shape)\n",
    "print(y_pred[0:15])\n",
    "print(y_test_t[0:15])\n",
    "y_pred_org = (y_pred * min_max_scaler.data_range_[3]) + min_max_scaler.data_min_[3] # min_max_scaler.inverse_transform(y_pred)\n",
    "y_test_t_org = (y_test_t * min_max_scaler.data_range_[3]) + min_max_scaler.data_min_[3] # min_max_scaler.inverse_transform(y_test_t)\n",
    "print(y_pred_org[0:15])\n",
    "print(y_test_t_org[0:15])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "def mean_absolute_percentage_error(y_test_t, y_pred): \n",
    "    y_test_t, y_pred = np.array(y_test_t), np.array(y_pred)\n",
    "    return np.mean(np.abs((y_test_t - y_pred) / y_test_t)) * 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4.485303878528074"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_absolute_percentage_error(y_test_t, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_time(text, stime):\n",
    "    seconds = (time.time()-stime)\n",
    "    print(text, seconds//60,\"minutes : \",np.round(seconds%60),\"seconds\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "program completed  8.0 minutes :  45.0 seconds\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the prediction\n",
    "from matplotlib import pyplot as plt\n",
    "#plt.figure()\n",
    "plt.figure(figsize=(10,5))\n",
    "plt.plot(y_pred_org)\n",
    "plt.plot(y_test_t_org)\n",
    "plt.title('LSTM Prediction vs Real Stock Price')\n",
    "plt.ylabel('Price')\n",
    "plt.xlabel('Days')\n",
    "plt.legend(['Prediction', 'Real'], loc='upper left')\n",
    "#plt.show()\n",
    "##plt.savefig(os.path.join(OUTPUT_PATH, 'pred_vs_real_BS'+str(BATCH_SIZE)+\"_\"+time.ctime()+'.png'))\n",
    "plt.savefig(os.path.join(OUTPUT_PATH, 'pred_vs_real_BS'+str(BATCH_SIZE)+\"_\"+time.strftime(\"%a %b %d %H %M %S %Y\", time.localtime())+'.png'))\n",
    "print_time(\"program completed \", stime)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}