LSTM_regression.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "8999ff1a",
   "metadata": {},
   "source": [
    "# Setup enviorment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "6dfc4cb6",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Data reading in Dataframe format and data preprocessing\n",
    "import pandas as pd\n",
    "from pandas import read_csv\n",
    "from pandas import DataFrame\n",
    "from pandas import concat\n",
    "\n",
    "# Data Visualization\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "# Linear algebra operations\n",
    "import numpy as np\n",
    "\n",
    "# Machine learning models and preprocessing\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import preprocessing\n",
    "\n",
    "# Deep learning\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from tensorflow.keras import Sequential, layers, callbacks\n",
    "from tensorflow.keras.layers import Dense, LSTM, Dropout, GRU, Bidirectional\n",
    "\n",
    "# Epiweek\n",
    "from epiweeks import Week, Year\n",
    "\n",
    "# Date\n",
    "from datetime import date as convert_to_date"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "0096f11d",
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "531f6bc1",
   "metadata": {},
   "outputs": [],
   "source": [
    "embeddings = 'Embeddings/embeddings_vae_1024features.csv'\n",
    "labels = 'Tabular_data/cases_per_inhabitants.csv'"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "05232315",
   "metadata": {},
   "source": [
    "# Read Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "26319b97",
   "metadata": {},
   "outputs": [],
   "source": [
    "def epiweek_from_date(image_date):\n",
    "    date = image_date.split('-')\n",
    "    \n",
    "    # Get year as int\n",
    "    year = ''.join(filter(str.isdigit, date[0]))\n",
    "    year = int(year)\n",
    "    \n",
    "    # Get month as int\n",
    "    month = ''.join(filter(str.isdigit, date[1]))\n",
    "    month = int(month)\n",
    "    \n",
    "    # Get day as int\n",
    "    day = ''.join(filter(str.isdigit, date[2]))\n",
    "    day = int(day)\n",
    "    \n",
    "    # Get epiweek:\n",
    "    date = convert_to_date(year, month, day)\n",
    "    epiweek = str(Week.fromdate(date))\n",
    "    epiweek = int(epiweek)\n",
    "    \n",
    "    return epiweek"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f184fe47",
   "metadata": {},
   "source": [
    "### 1. Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "0774267f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_features(path, Municipality = None):\n",
    "    df = pd.read_csv(path)\n",
    "    #df.Date = pd.to_datetime(df.Date)\n",
    "    \n",
    "    if Municipality:\n",
    "        print('Obtaining dataframe for the city of Medellin only...')\n",
    "        df = df[df['Municipality Code'] == Municipality]\n",
    "        \n",
    "    df.Date = df.Date.apply(epiweek_from_date)\n",
    "    \n",
    "    df = df.sort_values(by=['Date'])\n",
    "    \n",
    "    df = df.set_index('Date')\n",
    "    \n",
    "    if Municipality:\n",
    "        df.drop(columns=['Municipality Code'], inplace=True)\n",
    "        \n",
    "    df.index.name = None\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "3bdeed25",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Obtaining dataframe for the city of Medellin only...\n"
     ]
    },
    {
     "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>1014</th>\n",
       "      <th>1015</th>\n",
       "      <th>1016</th>\n",
       "      <th>1017</th>\n",
       "      <th>1018</th>\n",
       "      <th>1019</th>\n",
       "      <th>1020</th>\n",
       "      <th>1021</th>\n",
       "      <th>1022</th>\n",
       "      <th>1023</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>201544</th>\n",
       "      <td>-0.046489</td>\n",
       "      <td>-1.407201</td>\n",
       "      <td>-1.148883</td>\n",
       "      <td>0.836687</td>\n",
       "      <td>-0.190778</td>\n",
       "      <td>1.802231</td>\n",
       "      <td>0.517497</td>\n",
       "      <td>-1.146697</td>\n",
       "      <td>0.119171</td>\n",
       "      <td>1.544828</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.678144</td>\n",
       "      <td>-0.036498</td>\n",
       "      <td>-0.303533</td>\n",
       "      <td>0.248490</td>\n",
       "      <td>-0.811357</td>\n",
       "      <td>-0.832925</td>\n",
       "      <td>-0.739305</td>\n",
       "      <td>-0.295200</td>\n",
       "      <td>-0.038603</td>\n",
       "      <td>0.485674</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201545</th>\n",
       "      <td>0.327904</td>\n",
       "      <td>-0.701594</td>\n",
       "      <td>0.306197</td>\n",
       "      <td>-0.796750</td>\n",
       "      <td>-1.037395</td>\n",
       "      <td>0.857556</td>\n",
       "      <td>0.167456</td>\n",
       "      <td>1.642044</td>\n",
       "      <td>-0.473234</td>\n",
       "      <td>0.181270</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.897844</td>\n",
       "      <td>-0.661285</td>\n",
       "      <td>-1.146673</td>\n",
       "      <td>-1.234707</td>\n",
       "      <td>-0.184185</td>\n",
       "      <td>-0.140669</td>\n",
       "      <td>-2.367193</td>\n",
       "      <td>-0.463623</td>\n",
       "      <td>-0.668936</td>\n",
       "      <td>-0.233073</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201546</th>\n",
       "      <td>-1.634988</td>\n",
       "      <td>-0.882196</td>\n",
       "      <td>0.291420</td>\n",
       "      <td>-2.686286</td>\n",
       "      <td>-0.684004</td>\n",
       "      <td>-0.407912</td>\n",
       "      <td>1.827255</td>\n",
       "      <td>0.327352</td>\n",
       "      <td>0.987357</td>\n",
       "      <td>0.781570</td>\n",
       "      <td>...</td>\n",
       "      <td>0.906685</td>\n",
       "      <td>1.372330</td>\n",
       "      <td>-0.333867</td>\n",
       "      <td>0.728862</td>\n",
       "      <td>-0.005498</td>\n",
       "      <td>-0.743434</td>\n",
       "      <td>0.597319</td>\n",
       "      <td>0.195301</td>\n",
       "      <td>-0.850010</td>\n",
       "      <td>0.492581</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201547</th>\n",
       "      <td>1.789035</td>\n",
       "      <td>0.854698</td>\n",
       "      <td>-2.904163</td>\n",
       "      <td>-1.008464</td>\n",
       "      <td>-0.165688</td>\n",
       "      <td>0.597984</td>\n",
       "      <td>0.142369</td>\n",
       "      <td>0.164123</td>\n",
       "      <td>1.069377</td>\n",
       "      <td>0.236612</td>\n",
       "      <td>...</td>\n",
       "      <td>0.254610</td>\n",
       "      <td>-0.536157</td>\n",
       "      <td>-0.424962</td>\n",
       "      <td>0.685125</td>\n",
       "      <td>-0.029501</td>\n",
       "      <td>0.095485</td>\n",
       "      <td>0.379560</td>\n",
       "      <td>-0.297148</td>\n",
       "      <td>-1.211405</td>\n",
       "      <td>0.443391</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201548</th>\n",
       "      <td>-0.983180</td>\n",
       "      <td>0.145126</td>\n",
       "      <td>0.720849</td>\n",
       "      <td>1.524544</td>\n",
       "      <td>0.620878</td>\n",
       "      <td>-1.807810</td>\n",
       "      <td>0.340521</td>\n",
       "      <td>0.781281</td>\n",
       "      <td>-0.395229</td>\n",
       "      <td>-0.769379</td>\n",
       "      <td>...</td>\n",
       "      <td>1.562922</td>\n",
       "      <td>0.585608</td>\n",
       "      <td>-0.675990</td>\n",
       "      <td>0.232964</td>\n",
       "      <td>0.788353</td>\n",
       "      <td>0.147064</td>\n",
       "      <td>0.094748</td>\n",
       "      <td>-0.206415</td>\n",
       "      <td>1.170020</td>\n",
       "      <td>0.276934</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\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",
       "      <td>...</td>\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",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201848</th>\n",
       "      <td>-0.177821</td>\n",
       "      <td>0.800357</td>\n",
       "      <td>0.158949</td>\n",
       "      <td>2.573438</td>\n",
       "      <td>1.615337</td>\n",
       "      <td>-0.998008</td>\n",
       "      <td>1.096824</td>\n",
       "      <td>-0.023870</td>\n",
       "      <td>-0.667530</td>\n",
       "      <td>-1.929526</td>\n",
       "      <td>...</td>\n",
       "      <td>2.209477</td>\n",
       "      <td>0.241274</td>\n",
       "      <td>-1.311355</td>\n",
       "      <td>0.955204</td>\n",
       "      <td>0.006918</td>\n",
       "      <td>0.715542</td>\n",
       "      <td>-0.870650</td>\n",
       "      <td>-0.205803</td>\n",
       "      <td>0.835786</td>\n",
       "      <td>1.692061</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201849</th>\n",
       "      <td>-0.306444</td>\n",
       "      <td>1.308680</td>\n",
       "      <td>1.067536</td>\n",
       "      <td>-0.931025</td>\n",
       "      <td>-1.046062</td>\n",
       "      <td>2.075529</td>\n",
       "      <td>-0.501223</td>\n",
       "      <td>-2.267342</td>\n",
       "      <td>0.095569</td>\n",
       "      <td>-0.845871</td>\n",
       "      <td>...</td>\n",
       "      <td>-2.447275</td>\n",
       "      <td>0.293684</td>\n",
       "      <td>-1.791375</td>\n",
       "      <td>-0.086890</td>\n",
       "      <td>0.917567</td>\n",
       "      <td>0.631528</td>\n",
       "      <td>-2.004170</td>\n",
       "      <td>0.267849</td>\n",
       "      <td>1.151415</td>\n",
       "      <td>-1.051147</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201850</th>\n",
       "      <td>-0.151392</td>\n",
       "      <td>0.071449</td>\n",
       "      <td>0.356428</td>\n",
       "      <td>0.259774</td>\n",
       "      <td>-0.383608</td>\n",
       "      <td>1.065443</td>\n",
       "      <td>0.450759</td>\n",
       "      <td>0.322183</td>\n",
       "      <td>-1.307717</td>\n",
       "      <td>-0.181305</td>\n",
       "      <td>...</td>\n",
       "      <td>1.033319</td>\n",
       "      <td>0.951084</td>\n",
       "      <td>0.638319</td>\n",
       "      <td>-0.277678</td>\n",
       "      <td>0.032450</td>\n",
       "      <td>-0.329121</td>\n",
       "      <td>-0.322632</td>\n",
       "      <td>-1.898407</td>\n",
       "      <td>-0.602611</td>\n",
       "      <td>-0.267056</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201851</th>\n",
       "      <td>-0.793142</td>\n",
       "      <td>-0.539722</td>\n",
       "      <td>-1.165816</td>\n",
       "      <td>-0.199017</td>\n",
       "      <td>2.244849</td>\n",
       "      <td>0.917043</td>\n",
       "      <td>-0.467799</td>\n",
       "      <td>0.353362</td>\n",
       "      <td>1.053764</td>\n",
       "      <td>0.443968</td>\n",
       "      <td>...</td>\n",
       "      <td>0.882802</td>\n",
       "      <td>-0.827243</td>\n",
       "      <td>-0.127070</td>\n",
       "      <td>-1.308614</td>\n",
       "      <td>-1.632518</td>\n",
       "      <td>1.059085</td>\n",
       "      <td>1.047949</td>\n",
       "      <td>-2.762428</td>\n",
       "      <td>-0.955669</td>\n",
       "      <td>0.913588</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201852</th>\n",
       "      <td>-0.393870</td>\n",
       "      <td>1.455486</td>\n",
       "      <td>-0.128889</td>\n",
       "      <td>-0.788440</td>\n",
       "      <td>0.381879</td>\n",
       "      <td>0.022924</td>\n",
       "      <td>0.543938</td>\n",
       "      <td>-0.070586</td>\n",
       "      <td>-0.623824</td>\n",
       "      <td>-0.382842</td>\n",
       "      <td>...</td>\n",
       "      <td>0.705277</td>\n",
       "      <td>0.204254</td>\n",
       "      <td>-0.124967</td>\n",
       "      <td>0.008888</td>\n",
       "      <td>0.031512</td>\n",
       "      <td>2.580178</td>\n",
       "      <td>-0.813275</td>\n",
       "      <td>-0.418657</td>\n",
       "      <td>-0.572518</td>\n",
       "      <td>-0.936024</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>165 rows × 1024 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               0         1         2         3         4         5         6  \\\n",
       "201544 -0.046489 -1.407201 -1.148883  0.836687 -0.190778  1.802231  0.517497   \n",
       "201545  0.327904 -0.701594  0.306197 -0.796750 -1.037395  0.857556  0.167456   \n",
       "201546 -1.634988 -0.882196  0.291420 -2.686286 -0.684004 -0.407912  1.827255   \n",
       "201547  1.789035  0.854698 -2.904163 -1.008464 -0.165688  0.597984  0.142369   \n",
       "201548 -0.983180  0.145126  0.720849  1.524544  0.620878 -1.807810  0.340521   \n",
       "...          ...       ...       ...       ...       ...       ...       ...   \n",
       "201848 -0.177821  0.800357  0.158949  2.573438  1.615337 -0.998008  1.096824   \n",
       "201849 -0.306444  1.308680  1.067536 -0.931025 -1.046062  2.075529 -0.501223   \n",
       "201850 -0.151392  0.071449  0.356428  0.259774 -0.383608  1.065443  0.450759   \n",
       "201851 -0.793142 -0.539722 -1.165816 -0.199017  2.244849  0.917043 -0.467799   \n",
       "201852 -0.393870  1.455486 -0.128889 -0.788440  0.381879  0.022924  0.543938   \n",
       "\n",
       "               7         8         9  ...      1014      1015      1016  \\\n",
       "201544 -1.146697  0.119171  1.544828  ... -0.678144 -0.036498 -0.303533   \n",
       "201545  1.642044 -0.473234  0.181270  ... -0.897844 -0.661285 -1.146673   \n",
       "201546  0.327352  0.987357  0.781570  ...  0.906685  1.372330 -0.333867   \n",
       "201547  0.164123  1.069377  0.236612  ...  0.254610 -0.536157 -0.424962   \n",
       "201548  0.781281 -0.395229 -0.769379  ...  1.562922  0.585608 -0.675990   \n",
       "...          ...       ...       ...  ...       ...       ...       ...   \n",
       "201848 -0.023870 -0.667530 -1.929526  ...  2.209477  0.241274 -1.311355   \n",
       "201849 -2.267342  0.095569 -0.845871  ... -2.447275  0.293684 -1.791375   \n",
       "201850  0.322183 -1.307717 -0.181305  ...  1.033319  0.951084  0.638319   \n",
       "201851  0.353362  1.053764  0.443968  ...  0.882802 -0.827243 -0.127070   \n",
       "201852 -0.070586 -0.623824 -0.382842  ...  0.705277  0.204254 -0.124967   \n",
       "\n",
       "            1017      1018      1019      1020      1021      1022      1023  \n",
       "201544  0.248490 -0.811357 -0.832925 -0.739305 -0.295200 -0.038603  0.485674  \n",
       "201545 -1.234707 -0.184185 -0.140669 -2.367193 -0.463623 -0.668936 -0.233073  \n",
       "201546  0.728862 -0.005498 -0.743434  0.597319  0.195301 -0.850010  0.492581  \n",
       "201547  0.685125 -0.029501  0.095485  0.379560 -0.297148 -1.211405  0.443391  \n",
       "201548  0.232964  0.788353  0.147064  0.094748 -0.206415  1.170020  0.276934  \n",
       "...          ...       ...       ...       ...       ...       ...       ...  \n",
       "201848  0.955204  0.006918  0.715542 -0.870650 -0.205803  0.835786  1.692061  \n",
       "201849 -0.086890  0.917567  0.631528 -2.004170  0.267849  1.151415 -1.051147  \n",
       "201850 -0.277678  0.032450 -0.329121 -0.322632 -1.898407 -0.602611 -0.267056  \n",
       "201851 -1.308614 -1.632518  1.059085  1.047949 -2.762428 -0.955669  0.913588  \n",
       "201852  0.008888  0.031512  2.580178 -0.813275 -0.418657 -0.572518 -0.936024  \n",
       "\n",
       "[165 rows x 1024 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features_df = read_features(path=embeddings, Municipality='Medellín')\n",
    "features_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0486130d",
   "metadata": {},
   "source": [
    "### 2. Labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3cd05d7e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_epiweek(name):\n",
    "    \n",
    "    # Get week\n",
    "    week = name.split('/')[1]\n",
    "    week = week.replace('w','')\n",
    "    week = int(week)\n",
    "    \n",
    "    # Year\n",
    "    year = name.split('/')[0]\n",
    "    year = int(year)\n",
    "    \n",
    "    epiweek = Week(year, week)\n",
    "    \n",
    "    epiweek = str(epiweek)\n",
    "    epiweek = int(epiweek)\n",
    "\n",
    "    return epiweek"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "f354c0b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_labels(path, Municipality = None):\n",
    "    df = pd.read_csv(path)\n",
    "    if df.shape[1] > 678:\n",
    "        df = pd.concat([df[['Municipality code', 'Municipality']], df.iloc[:,-676:]], axis=1)\n",
    "        cols = df.iloc[:, 2:].columns\n",
    "        new_cols = df.iloc[:, 2:].columns.to_series().apply(get_epiweek)\n",
    "        df = df.rename(columns=dict(zip(cols, new_cols))) \n",
    "        \n",
    "    if 'Label_CSV_All_Municipality' in path:\n",
    "        # Get Columns\n",
    "        df = df[['epiweek', 'Municipality code', 'Municipality', 'final_cases_label']]\n",
    "        \n",
    "        # change epiweek format\n",
    "        df.epiweek = df.epiweek.apply(get_epiweek)\n",
    "        \n",
    "        # Remove duplicates\n",
    "        df = df[df.duplicated(['epiweek','Municipality code','Municipality']) == False]\n",
    "        \n",
    "        # Replace Increase, decrease, stable to numerical:\n",
    "        \"\"\"\n",
    "        - Stable = 0\n",
    "        - Increased = 1 \n",
    "        - Decreased = 2\n",
    "        \"\"\"\n",
    "        df.final_cases_label = df.final_cases_label.replace({'Stable': 0, 'Increased': 1, 'Decreased': 2})\n",
    "        \n",
    "        # Create table\n",
    "        df = df.pivot(index=['Municipality code', 'Municipality'], columns='epiweek', values='final_cases_label')\n",
    "\n",
    "        # Reset Index:\n",
    "        df = df.reset_index()\n",
    "    \n",
    "    if Municipality:\n",
    "        df = df[df['Municipality'] == Municipality]\n",
    "        df.drop(columns=['Municipality code'], inplace=True)\n",
    "        df.rename(columns={'Municipality': 'Municipality Code'}, inplace=True)\n",
    "    \n",
    "        df = df.set_index('Municipality Code')\n",
    "        df = df.T\n",
    "\n",
    "        df.columns.name = None\n",
    "        df.index.name = None\n",
    "        \n",
    "        df.columns = ['Labels']\n",
    "        \n",
    "        df.index = pd.to_numeric(df.index)\n",
    "    \n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "e2cdd788",
   "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>Labels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>200701</th>\n",
       "      <td>4.374212e-07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>200702</th>\n",
       "      <td>0.000000e+00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>200703</th>\n",
       "      <td>0.000000e+00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>200704</th>\n",
       "      <td>0.000000e+00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>200705</th>\n",
       "      <td>0.000000e+00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201948</th>\n",
       "      <td>6.039754e-06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201949</th>\n",
       "      <td>8.053005e-06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201950</th>\n",
       "      <td>1.207951e-05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201951</th>\n",
       "      <td>5.637103e-06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201952</th>\n",
       "      <td>2.013251e-06</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>676 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              Labels\n",
       "200701  4.374212e-07\n",
       "200702  0.000000e+00\n",
       "200703  0.000000e+00\n",
       "200704  0.000000e+00\n",
       "200705  0.000000e+00\n",
       "...              ...\n",
       "201948  6.039754e-06\n",
       "201949  8.053005e-06\n",
       "201950  1.207951e-05\n",
       "201951  5.637103e-06\n",
       "201952  2.013251e-06\n",
       "\n",
       "[676 rows x 1 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels_df = read_labels(path=labels, Municipality='Medellín')\n",
    "labels_df "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "06e126cb",
   "metadata": {},
   "source": [
    "# Data preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "811edfa0",
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>1015</th>\n",
       "      <th>1016</th>\n",
       "      <th>1017</th>\n",
       "      <th>1018</th>\n",
       "      <th>1019</th>\n",
       "      <th>1020</th>\n",
       "      <th>1021</th>\n",
       "      <th>1022</th>\n",
       "      <th>1023</th>\n",
       "      <th>Labels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>201544</th>\n",
       "      <td>-0.046489</td>\n",
       "      <td>-1.407201</td>\n",
       "      <td>-1.148883</td>\n",
       "      <td>0.836687</td>\n",
       "      <td>-0.190778</td>\n",
       "      <td>1.802231</td>\n",
       "      <td>0.517497</td>\n",
       "      <td>-1.146697</td>\n",
       "      <td>0.119171</td>\n",
       "      <td>1.544828</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.036498</td>\n",
       "      <td>-0.303533</td>\n",
       "      <td>0.248490</td>\n",
       "      <td>-0.811357</td>\n",
       "      <td>-0.832925</td>\n",
       "      <td>-0.739305</td>\n",
       "      <td>-0.295200</td>\n",
       "      <td>-0.038603</td>\n",
       "      <td>0.485674</td>\n",
       "      <td>0.000042</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201545</th>\n",
       "      <td>0.327904</td>\n",
       "      <td>-0.701594</td>\n",
       "      <td>0.306197</td>\n",
       "      <td>-0.796750</td>\n",
       "      <td>-1.037395</td>\n",
       "      <td>0.857556</td>\n",
       "      <td>0.167456</td>\n",
       "      <td>1.642044</td>\n",
       "      <td>-0.473234</td>\n",
       "      <td>0.181270</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.661285</td>\n",
       "      <td>-1.146673</td>\n",
       "      <td>-1.234707</td>\n",
       "      <td>-0.184185</td>\n",
       "      <td>-0.140669</td>\n",
       "      <td>-2.367193</td>\n",
       "      <td>-0.463623</td>\n",
       "      <td>-0.668936</td>\n",
       "      <td>-0.233073</td>\n",
       "      <td>0.000050</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201546</th>\n",
       "      <td>-1.634988</td>\n",
       "      <td>-0.882196</td>\n",
       "      <td>0.291420</td>\n",
       "      <td>-2.686286</td>\n",
       "      <td>-0.684004</td>\n",
       "      <td>-0.407912</td>\n",
       "      <td>1.827255</td>\n",
       "      <td>0.327352</td>\n",
       "      <td>0.987357</td>\n",
       "      <td>0.781570</td>\n",
       "      <td>...</td>\n",
       "      <td>1.372330</td>\n",
       "      <td>-0.333867</td>\n",
       "      <td>0.728862</td>\n",
       "      <td>-0.005498</td>\n",
       "      <td>-0.743434</td>\n",
       "      <td>0.597319</td>\n",
       "      <td>0.195301</td>\n",
       "      <td>-0.850010</td>\n",
       "      <td>0.492581</td>\n",
       "      <td>0.000043</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201547</th>\n",
       "      <td>1.789035</td>\n",
       "      <td>0.854698</td>\n",
       "      <td>-2.904163</td>\n",
       "      <td>-1.008464</td>\n",
       "      <td>-0.165688</td>\n",
       "      <td>0.597984</td>\n",
       "      <td>0.142369</td>\n",
       "      <td>0.164123</td>\n",
       "      <td>1.069377</td>\n",
       "      <td>0.236612</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.536157</td>\n",
       "      <td>-0.424962</td>\n",
       "      <td>0.685125</td>\n",
       "      <td>-0.029501</td>\n",
       "      <td>0.095485</td>\n",
       "      <td>0.379560</td>\n",
       "      <td>-0.297148</td>\n",
       "      <td>-1.211405</td>\n",
       "      <td>0.443391</td>\n",
       "      <td>0.000040</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201548</th>\n",
       "      <td>-0.983180</td>\n",
       "      <td>0.145126</td>\n",
       "      <td>0.720849</td>\n",
       "      <td>1.524544</td>\n",
       "      <td>0.620878</td>\n",
       "      <td>-1.807810</td>\n",
       "      <td>0.340521</td>\n",
       "      <td>0.781281</td>\n",
       "      <td>-0.395229</td>\n",
       "      <td>-0.769379</td>\n",
       "      <td>...</td>\n",
       "      <td>0.585608</td>\n",
       "      <td>-0.675990</td>\n",
       "      <td>0.232964</td>\n",
       "      <td>0.788353</td>\n",
       "      <td>0.147064</td>\n",
       "      <td>0.094748</td>\n",
       "      <td>-0.206415</td>\n",
       "      <td>1.170020</td>\n",
       "      <td>0.276934</td>\n",
       "      <td>0.000056</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\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",
       "      <td>...</td>\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",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201848</th>\n",
       "      <td>-0.177821</td>\n",
       "      <td>0.800357</td>\n",
       "      <td>0.158949</td>\n",
       "      <td>2.573438</td>\n",
       "      <td>1.615337</td>\n",
       "      <td>-0.998008</td>\n",
       "      <td>1.096824</td>\n",
       "      <td>-0.023870</td>\n",
       "      <td>-0.667530</td>\n",
       "      <td>-1.929526</td>\n",
       "      <td>...</td>\n",
       "      <td>0.241274</td>\n",
       "      <td>-1.311355</td>\n",
       "      <td>0.955204</td>\n",
       "      <td>0.006918</td>\n",
       "      <td>0.715542</td>\n",
       "      <td>-0.870650</td>\n",
       "      <td>-0.205803</td>\n",
       "      <td>0.835786</td>\n",
       "      <td>1.692061</td>\n",
       "      <td>0.000012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201849</th>\n",
       "      <td>-0.306444</td>\n",
       "      <td>1.308680</td>\n",
       "      <td>1.067536</td>\n",
       "      <td>-0.931025</td>\n",
       "      <td>-1.046062</td>\n",
       "      <td>2.075529</td>\n",
       "      <td>-0.501223</td>\n",
       "      <td>-2.267342</td>\n",
       "      <td>0.095569</td>\n",
       "      <td>-0.845871</td>\n",
       "      <td>...</td>\n",
       "      <td>0.293684</td>\n",
       "      <td>-1.791375</td>\n",
       "      <td>-0.086890</td>\n",
       "      <td>0.917567</td>\n",
       "      <td>0.631528</td>\n",
       "      <td>-2.004170</td>\n",
       "      <td>0.267849</td>\n",
       "      <td>1.151415</td>\n",
       "      <td>-1.051147</td>\n",
       "      <td>0.000005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201850</th>\n",
       "      <td>-0.151392</td>\n",
       "      <td>0.071449</td>\n",
       "      <td>0.356428</td>\n",
       "      <td>0.259774</td>\n",
       "      <td>-0.383608</td>\n",
       "      <td>1.065443</td>\n",
       "      <td>0.450759</td>\n",
       "      <td>0.322183</td>\n",
       "      <td>-1.307717</td>\n",
       "      <td>-0.181305</td>\n",
       "      <td>...</td>\n",
       "      <td>0.951084</td>\n",
       "      <td>0.638319</td>\n",
       "      <td>-0.277678</td>\n",
       "      <td>0.032450</td>\n",
       "      <td>-0.329121</td>\n",
       "      <td>-0.322632</td>\n",
       "      <td>-1.898407</td>\n",
       "      <td>-0.602611</td>\n",
       "      <td>-0.267056</td>\n",
       "      <td>0.000011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201851</th>\n",
       "      <td>-0.793142</td>\n",
       "      <td>-0.539722</td>\n",
       "      <td>-1.165816</td>\n",
       "      <td>-0.199017</td>\n",
       "      <td>2.244849</td>\n",
       "      <td>0.917043</td>\n",
       "      <td>-0.467799</td>\n",
       "      <td>0.353362</td>\n",
       "      <td>1.053764</td>\n",
       "      <td>0.443968</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.827243</td>\n",
       "      <td>-0.127070</td>\n",
       "      <td>-1.308614</td>\n",
       "      <td>-1.632518</td>\n",
       "      <td>1.059085</td>\n",
       "      <td>1.047949</td>\n",
       "      <td>-2.762428</td>\n",
       "      <td>-0.955669</td>\n",
       "      <td>0.913588</td>\n",
       "      <td>0.000007</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201852</th>\n",
       "      <td>-0.393870</td>\n",
       "      <td>1.455486</td>\n",
       "      <td>-0.128889</td>\n",
       "      <td>-0.788440</td>\n",
       "      <td>0.381879</td>\n",
       "      <td>0.022924</td>\n",
       "      <td>0.543938</td>\n",
       "      <td>-0.070586</td>\n",
       "      <td>-0.623824</td>\n",
       "      <td>-0.382842</td>\n",
       "      <td>...</td>\n",
       "      <td>0.204254</td>\n",
       "      <td>-0.124967</td>\n",
       "      <td>0.008888</td>\n",
       "      <td>0.031512</td>\n",
       "      <td>2.580178</td>\n",
       "      <td>-0.813275</td>\n",
       "      <td>-0.418657</td>\n",
       "      <td>-0.572518</td>\n",
       "      <td>-0.936024</td>\n",
       "      <td>0.000007</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>165 rows × 1025 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               0         1         2         3         4         5         6  \\\n",
       "201544 -0.046489 -1.407201 -1.148883  0.836687 -0.190778  1.802231  0.517497   \n",
       "201545  0.327904 -0.701594  0.306197 -0.796750 -1.037395  0.857556  0.167456   \n",
       "201546 -1.634988 -0.882196  0.291420 -2.686286 -0.684004 -0.407912  1.827255   \n",
       "201547  1.789035  0.854698 -2.904163 -1.008464 -0.165688  0.597984  0.142369   \n",
       "201548 -0.983180  0.145126  0.720849  1.524544  0.620878 -1.807810  0.340521   \n",
       "...          ...       ...       ...       ...       ...       ...       ...   \n",
       "201848 -0.177821  0.800357  0.158949  2.573438  1.615337 -0.998008  1.096824   \n",
       "201849 -0.306444  1.308680  1.067536 -0.931025 -1.046062  2.075529 -0.501223   \n",
       "201850 -0.151392  0.071449  0.356428  0.259774 -0.383608  1.065443  0.450759   \n",
       "201851 -0.793142 -0.539722 -1.165816 -0.199017  2.244849  0.917043 -0.467799   \n",
       "201852 -0.393870  1.455486 -0.128889 -0.788440  0.381879  0.022924  0.543938   \n",
       "\n",
       "               7         8         9  ...      1015      1016      1017  \\\n",
       "201544 -1.146697  0.119171  1.544828  ... -0.036498 -0.303533  0.248490   \n",
       "201545  1.642044 -0.473234  0.181270  ... -0.661285 -1.146673 -1.234707   \n",
       "201546  0.327352  0.987357  0.781570  ...  1.372330 -0.333867  0.728862   \n",
       "201547  0.164123  1.069377  0.236612  ... -0.536157 -0.424962  0.685125   \n",
       "201548  0.781281 -0.395229 -0.769379  ...  0.585608 -0.675990  0.232964   \n",
       "...          ...       ...       ...  ...       ...       ...       ...   \n",
       "201848 -0.023870 -0.667530 -1.929526  ...  0.241274 -1.311355  0.955204   \n",
       "201849 -2.267342  0.095569 -0.845871  ...  0.293684 -1.791375 -0.086890   \n",
       "201850  0.322183 -1.307717 -0.181305  ...  0.951084  0.638319 -0.277678   \n",
       "201851  0.353362  1.053764  0.443968  ... -0.827243 -0.127070 -1.308614   \n",
       "201852 -0.070586 -0.623824 -0.382842  ...  0.204254 -0.124967  0.008888   \n",
       "\n",
       "            1018      1019      1020      1021      1022      1023    Labels  \n",
       "201544 -0.811357 -0.832925 -0.739305 -0.295200 -0.038603  0.485674  0.000042  \n",
       "201545 -0.184185 -0.140669 -2.367193 -0.463623 -0.668936 -0.233073  0.000050  \n",
       "201546 -0.005498 -0.743434  0.597319  0.195301 -0.850010  0.492581  0.000043  \n",
       "201547 -0.029501  0.095485  0.379560 -0.297148 -1.211405  0.443391  0.000040  \n",
       "201548  0.788353  0.147064  0.094748 -0.206415  1.170020  0.276934  0.000056  \n",
       "...          ...       ...       ...       ...       ...       ...       ...  \n",
       "201848  0.006918  0.715542 -0.870650 -0.205803  0.835786  1.692061  0.000012  \n",
       "201849  0.917567  0.631528 -2.004170  0.267849  1.151415 -1.051147  0.000005  \n",
       "201850  0.032450 -0.329121 -0.322632 -1.898407 -0.602611 -0.267056  0.000011  \n",
       "201851 -1.632518  1.059085  1.047949 -2.762428 -0.955669  0.913588  0.000007  \n",
       "201852  0.031512  2.580178 -0.813275 -0.418657 -0.572518 -0.936024  0.000007  \n",
       "\n",
       "[165 rows x 1025 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Merge the two dataframes based on the date values\n",
    "dengue_df = features_df.merge(labels_df, how='inner', left_index=True, right_index=True)\n",
    "dengue_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b6cda4c4",
   "metadata": {},
   "source": [
    "### Train Test split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "e2965d39",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_test_split(df, train_percentage = 80):\n",
    "    # We need a sequence so we can't split randomly\n",
    "    # To divide into Train and test we have to calculate the train percentage of the dataset:\n",
    "    size = df.shape[0]\n",
    "    split = int(size*(train_percentage/100))\n",
    "    \n",
    "    \"\"\" Train \"\"\"\n",
    "    # We will train with 1st percentage % of data and test with the rest\n",
    "    train_df = df.iloc[:split,:] ## percentage % train\n",
    "    \n",
    "    \"\"\" Test \"\"\"\n",
    "    test_df = df.iloc[split:,:] # 100 - percentage % test\n",
    "    \n",
    "    print(f'The train shape is: {train_df.shape}')\n",
    "    print(f'The test shape is: {test_df.shape}')\n",
    "    \n",
    "    return train_df, test_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "6366a512",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The train shape is: (132, 1025)\n",
      "The test shape is: (33, 1025)\n"
     ]
    }
   ],
   "source": [
    "train_df, test_df = train_test_split(dengue_df, train_percentage = 80)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6ae92ccc",
   "metadata": {},
   "source": [
    "### Normalize features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "673b6ecb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Normalize train data and create the scaler\n",
    "def normalize_train_features(df, feature_range=(-1, 1)):\n",
    "    \n",
    "    scalers = {}\n",
    "    # For each column in the dataframe\n",
    "    for column in df.columns:\n",
    "        # Get values of the column\n",
    "        values = df[column].values.reshape(-1,1)\n",
    "        # Generate a new scaler\n",
    "        scaler = MinMaxScaler(feature_range=feature_range)\n",
    "        # Fit the scaler just for that column\n",
    "        scaled_column = scaler.fit_transform(values)\n",
    "        # Add the scaled column to the dataframe\n",
    "        scaled_column = np.reshape(scaled_column, len(scaled_column))\n",
    "        df[column] = scaled_column\n",
    "        \n",
    "        # Save the scaler of the column\n",
    "        scalers['scaler_' + column] = scaler\n",
    "        \n",
    "    print(f' Min values are: ')\n",
    "    print(df.min())\n",
    "    print(f' Max values are: ')\n",
    "    print(df.max())\n",
    "        \n",
    "    return df, scalers\n",
    "\n",
    "\n",
    "\"\"\" If you want to use the same scaler used in train, you can use this function\"\"\"\n",
    "def normalize_test_features(df, scalers=None):\n",
    "    \n",
    "    if not scalers:\n",
    "        raise TypeError(\"You should provide a list of scalers.\")\n",
    "        \n",
    "    for column in df.columns:\n",
    "        # Get values of the column\n",
    "        values = df[column].values.reshape(-1,1)\n",
    "        # Take the scaler of that column\n",
    "        scaler = scalers['scaler_' + column]\n",
    "        # Scale values\n",
    "        scaled_column = scaler.transform(values)\n",
    "        scaled_column = np.reshape(scaled_column,len(scaled_column))\n",
    "        # Add the scaled values to the df\n",
    "        df[column] = scaled_column\n",
    "        \n",
    "    print(f' Min values are: ')\n",
    "    print(df.min())\n",
    "    print(f' Max values are: ')\n",
    "    print(df.max())\n",
    "        \n",
    "    return df "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "3f76b493",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Min values are: \n",
      "0        -1.0\n",
      "1        -1.0\n",
      "2        -1.0\n",
      "3        -1.0\n",
      "4        -1.0\n",
      "         ... \n",
      "1020     -1.0\n",
      "1021     -1.0\n",
      "1022     -1.0\n",
      "1023     -1.0\n",
      "Labels   -1.0\n",
      "Length: 1025, dtype: float64\n",
      " Max values are: \n",
      "0         1.0\n",
      "1         1.0\n",
      "2         1.0\n",
      "3         1.0\n",
      "4         1.0\n",
      "         ... \n",
      "1020      1.0\n",
      "1021      1.0\n",
      "1022      1.0\n",
      "1023      1.0\n",
      "Labels    1.0\n",
      "Length: 1025, dtype: float64\n"
     ]
    },
    {
     "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>1015</th>\n",
       "      <th>1016</th>\n",
       "      <th>1017</th>\n",
       "      <th>1018</th>\n",
       "      <th>1019</th>\n",
       "      <th>1020</th>\n",
       "      <th>1021</th>\n",
       "      <th>1022</th>\n",
       "      <th>1023</th>\n",
       "      <th>Labels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>201544</th>\n",
       "      <td>0.153002</td>\n",
       "      <td>-0.545807</td>\n",
       "      <td>-0.370448</td>\n",
       "      <td>0.304693</td>\n",
       "      <td>-0.024118</td>\n",
       "      <td>0.540703</td>\n",
       "      <td>0.174470</td>\n",
       "      <td>-0.667477</td>\n",
       "      <td>0.235845</td>\n",
       "      <td>0.535753</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.166016</td>\n",
       "      <td>-0.019667</td>\n",
       "      <td>0.183710</td>\n",
       "      <td>-0.527411</td>\n",
       "      <td>-0.256853</td>\n",
       "      <td>-0.149288</td>\n",
       "      <td>0.025600</td>\n",
       "      <td>-0.013572</td>\n",
       "      <td>0.242906</td>\n",
       "      <td>-0.693564</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201545</th>\n",
       "      <td>0.311409</td>\n",
       "      <td>-0.269567</td>\n",
       "      <td>0.151433</td>\n",
       "      <td>-0.300232</td>\n",
       "      <td>-0.319257</td>\n",
       "      <td>0.187638</td>\n",
       "      <td>0.025189</td>\n",
       "      <td>0.388860</td>\n",
       "      <td>0.036709</td>\n",
       "      <td>-0.028918</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.370045</td>\n",
       "      <td>-0.351644</td>\n",
       "      <td>-0.495220</td>\n",
       "      <td>-0.251282</td>\n",
       "      <td>-0.019697</td>\n",
       "      <td>-0.720239</td>\n",
       "      <td>-0.028609</td>\n",
       "      <td>-0.232408</td>\n",
       "      <td>-0.078399</td>\n",
       "      <td>-0.631570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201546</th>\n",
       "      <td>-0.519102</td>\n",
       "      <td>-0.340271</td>\n",
       "      <td>0.146133</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>-0.196062</td>\n",
       "      <td>-0.285319</td>\n",
       "      <td>0.733040</td>\n",
       "      <td>-0.109127</td>\n",
       "      <td>0.527685</td>\n",
       "      <td>0.219675</td>\n",
       "      <td>...</td>\n",
       "      <td>0.294045</td>\n",
       "      <td>-0.031611</td>\n",
       "      <td>0.403600</td>\n",
       "      <td>-0.172611</td>\n",
       "      <td>-0.226195</td>\n",
       "      <td>0.319509</td>\n",
       "      <td>0.183477</td>\n",
       "      <td>-0.295272</td>\n",
       "      <td>0.245994</td>\n",
       "      <td>-0.686676</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201547</th>\n",
       "      <td>0.929623</td>\n",
       "      <td>0.339709</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>-0.378638</td>\n",
       "      <td>-0.015372</td>\n",
       "      <td>0.090626</td>\n",
       "      <td>0.014490</td>\n",
       "      <td>-0.170956</td>\n",
       "      <td>0.555255</td>\n",
       "      <td>-0.006000</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.329183</td>\n",
       "      <td>-0.067479</td>\n",
       "      <td>0.383579</td>\n",
       "      <td>-0.183179</td>\n",
       "      <td>0.061207</td>\n",
       "      <td>0.243134</td>\n",
       "      <td>0.024973</td>\n",
       "      <td>-0.420739</td>\n",
       "      <td>0.224004</td>\n",
       "      <td>-0.714229</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201548</th>\n",
       "      <td>-0.243318</td>\n",
       "      <td>0.061918</td>\n",
       "      <td>0.300152</td>\n",
       "      <td>0.559433</td>\n",
       "      <td>0.258833</td>\n",
       "      <td>-0.808519</td>\n",
       "      <td>0.098995</td>\n",
       "      <td>0.062815</td>\n",
       "      <td>0.062930</td>\n",
       "      <td>-0.422597</td>\n",
       "      <td>...</td>\n",
       "      <td>0.037136</td>\n",
       "      <td>-0.166318</td>\n",
       "      <td>0.176603</td>\n",
       "      <td>0.176902</td>\n",
       "      <td>0.078877</td>\n",
       "      <td>0.143241</td>\n",
       "      <td>0.054177</td>\n",
       "      <td>0.406031</td>\n",
       "      <td>0.149592</td>\n",
       "      <td>-0.583352</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 1025 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               0         1         2         3         4         5         6  \\\n",
       "201544  0.153002 -0.545807 -0.370448  0.304693 -0.024118  0.540703  0.174470   \n",
       "201545  0.311409 -0.269567  0.151433 -0.300232 -0.319257  0.187638  0.025189   \n",
       "201546 -0.519102 -0.340271  0.146133 -1.000000 -0.196062 -0.285319  0.733040   \n",
       "201547  0.929623  0.339709 -1.000000 -0.378638 -0.015372  0.090626  0.014490   \n",
       "201548 -0.243318  0.061918  0.300152  0.559433  0.258833 -0.808519  0.098995   \n",
       "\n",
       "               7         8         9  ...      1015      1016      1017  \\\n",
       "201544 -0.667477  0.235845  0.535753  ... -0.166016 -0.019667  0.183710   \n",
       "201545  0.388860  0.036709 -0.028918  ... -0.370045 -0.351644 -0.495220   \n",
       "201546 -0.109127  0.527685  0.219675  ...  0.294045 -0.031611  0.403600   \n",
       "201547 -0.170956  0.555255 -0.006000  ... -0.329183 -0.067479  0.383579   \n",
       "201548  0.062815  0.062930 -0.422597  ...  0.037136 -0.166318  0.176603   \n",
       "\n",
       "            1018      1019      1020      1021      1022      1023    Labels  \n",
       "201544 -0.527411 -0.256853 -0.149288  0.025600 -0.013572  0.242906 -0.693564  \n",
       "201545 -0.251282 -0.019697 -0.720239 -0.028609 -0.232408 -0.078399 -0.631570  \n",
       "201546 -0.172611 -0.226195  0.319509  0.183477 -0.295272  0.245994 -0.686676  \n",
       "201547 -0.183179  0.061207  0.243134  0.024973 -0.420739  0.224004 -0.714229  \n",
       "201548  0.176902  0.078877  0.143241  0.054177  0.406031  0.149592 -0.583352  \n",
       "\n",
       "[5 rows x 1025 columns]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_range = (-1, 1)\n",
    "\n",
    "# Scale train:\n",
    "train_df, scalers = normalize_train_features(train_df, feature_range=feature_range)\n",
    "\n",
    "#print(f'The scalers are: {scalers}')\n",
    "\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "299edd5f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Min values are: \n",
      "0        -0.611774\n",
      "1        -0.698647\n",
      "2        -1.048959\n",
      "3        -0.447103\n",
      "4        -0.736973\n",
      "            ...   \n",
      "1020     -0.645408\n",
      "1021     -0.768519\n",
      "1022     -0.468700\n",
      "1023     -1.009303\n",
      "Labels   -1.006901\n",
      "Length: 1025, dtype: float64\n",
      " Max values are: \n",
      "0         0.798543\n",
      "1         0.992657\n",
      "2         0.487560\n",
      "3         0.947879\n",
      "4         0.824965\n",
      "            ...   \n",
      "1020      0.705626\n",
      "1021      0.787044\n",
      "1022      0.451219\n",
      "1023      0.782204\n",
      "Labels   -0.906835\n",
      "Length: 1025, dtype: float64\n"
     ]
    },
    {
     "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>1015</th>\n",
       "      <th>1016</th>\n",
       "      <th>1017</th>\n",
       "      <th>1018</th>\n",
       "      <th>1019</th>\n",
       "      <th>1020</th>\n",
       "      <th>1021</th>\n",
       "      <th>1022</th>\n",
       "      <th>1023</th>\n",
       "      <th>Labels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>201820</th>\n",
       "      <td>0.151612</td>\n",
       "      <td>0.610249</td>\n",
       "      <td>0.437266</td>\n",
       "      <td>-0.337245</td>\n",
       "      <td>-0.736973</td>\n",
       "      <td>0.029515</td>\n",
       "      <td>-0.034929</td>\n",
       "      <td>-0.228096</td>\n",
       "      <td>0.653207</td>\n",
       "      <td>0.540979</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.376762</td>\n",
       "      <td>-0.328220</td>\n",
       "      <td>0.502141</td>\n",
       "      <td>-0.403143</td>\n",
       "      <td>-0.062169</td>\n",
       "      <td>-0.114325</td>\n",
       "      <td>-0.087839</td>\n",
       "      <td>0.094620</td>\n",
       "      <td>-0.479982</td>\n",
       "      <td>-0.975846</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201821</th>\n",
       "      <td>0.084599</td>\n",
       "      <td>0.542881</td>\n",
       "      <td>-0.137862</td>\n",
       "      <td>0.008561</td>\n",
       "      <td>0.138302</td>\n",
       "      <td>0.078038</td>\n",
       "      <td>-0.007716</td>\n",
       "      <td>-0.031384</td>\n",
       "      <td>-0.545694</td>\n",
       "      <td>-0.072470</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.445808</td>\n",
       "      <td>-0.618029</td>\n",
       "      <td>-0.699482</td>\n",
       "      <td>0.136318</td>\n",
       "      <td>0.328906</td>\n",
       "      <td>0.342614</td>\n",
       "      <td>0.063277</td>\n",
       "      <td>0.219823</td>\n",
       "      <td>0.294464</td>\n",
       "      <td>-0.989648</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201822</th>\n",
       "      <td>0.798543</td>\n",
       "      <td>0.992657</td>\n",
       "      <td>-0.227682</td>\n",
       "      <td>-0.299051</td>\n",
       "      <td>0.266510</td>\n",
       "      <td>-0.591903</td>\n",
       "      <td>0.565659</td>\n",
       "      <td>-0.179126</td>\n",
       "      <td>0.237211</td>\n",
       "      <td>0.382436</td>\n",
       "      <td>...</td>\n",
       "      <td>0.249908</td>\n",
       "      <td>-0.116217</td>\n",
       "      <td>0.201944</td>\n",
       "      <td>-0.673409</td>\n",
       "      <td>-0.036940</td>\n",
       "      <td>-0.021246</td>\n",
       "      <td>0.060466</td>\n",
       "      <td>0.102064</td>\n",
       "      <td>0.112100</td>\n",
       "      <td>-0.982747</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201823</th>\n",
       "      <td>0.247549</td>\n",
       "      <td>-0.247999</td>\n",
       "      <td>0.386614</td>\n",
       "      <td>0.358396</td>\n",
       "      <td>0.187533</td>\n",
       "      <td>-0.202666</td>\n",
       "      <td>-0.067352</td>\n",
       "      <td>-0.504047</td>\n",
       "      <td>0.361369</td>\n",
       "      <td>0.378615</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.429229</td>\n",
       "      <td>-0.519579</td>\n",
       "      <td>-0.290135</td>\n",
       "      <td>0.185387</td>\n",
       "      <td>0.484507</td>\n",
       "      <td>-0.645408</td>\n",
       "      <td>-0.595917</td>\n",
       "      <td>-0.186724</td>\n",
       "      <td>-0.197858</td>\n",
       "      <td>-0.968945</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201824</th>\n",
       "      <td>0.420624</td>\n",
       "      <td>0.373608</td>\n",
       "      <td>-0.479165</td>\n",
       "      <td>0.551768</td>\n",
       "      <td>-0.279726</td>\n",
       "      <td>-0.603021</td>\n",
       "      <td>0.154349</td>\n",
       "      <td>-0.543294</td>\n",
       "      <td>0.365547</td>\n",
       "      <td>0.051451</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.589635</td>\n",
       "      <td>-0.518102</td>\n",
       "      <td>-0.053085</td>\n",
       "      <td>-0.549938</td>\n",
       "      <td>0.159301</td>\n",
       "      <td>0.492040</td>\n",
       "      <td>-0.004100</td>\n",
       "      <td>0.447210</td>\n",
       "      <td>-0.103205</td>\n",
       "      <td>-0.972396</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 1025 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               0         1         2         3         4         5         6  \\\n",
       "201820  0.151612  0.610249  0.437266 -0.337245 -0.736973  0.029515 -0.034929   \n",
       "201821  0.084599  0.542881 -0.137862  0.008561  0.138302  0.078038 -0.007716   \n",
       "201822  0.798543  0.992657 -0.227682 -0.299051  0.266510 -0.591903  0.565659   \n",
       "201823  0.247549 -0.247999  0.386614  0.358396  0.187533 -0.202666 -0.067352   \n",
       "201824  0.420624  0.373608 -0.479165  0.551768 -0.279726 -0.603021  0.154349   \n",
       "\n",
       "               7         8         9  ...      1015      1016      1017  \\\n",
       "201820 -0.228096  0.653207  0.540979  ... -0.376762 -0.328220  0.502141   \n",
       "201821 -0.031384 -0.545694 -0.072470  ... -0.445808 -0.618029 -0.699482   \n",
       "201822 -0.179126  0.237211  0.382436  ...  0.249908 -0.116217  0.201944   \n",
       "201823 -0.504047  0.361369  0.378615  ... -0.429229 -0.519579 -0.290135   \n",
       "201824 -0.543294  0.365547  0.051451  ... -0.589635 -0.518102 -0.053085   \n",
       "\n",
       "            1018      1019      1020      1021      1022      1023    Labels  \n",
       "201820 -0.403143 -0.062169 -0.114325 -0.087839  0.094620 -0.479982 -0.975846  \n",
       "201821  0.136318  0.328906  0.342614  0.063277  0.219823  0.294464 -0.989648  \n",
       "201822 -0.673409 -0.036940 -0.021246  0.060466  0.102064  0.112100 -0.982747  \n",
       "201823  0.185387  0.484507 -0.645408 -0.595917 -0.186724 -0.197858 -0.968945  \n",
       "201824 -0.549938  0.159301  0.492040 -0.004100  0.447210 -0.103205 -0.972396  \n",
       "\n",
       "[5 rows x 1025 columns]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Scale test:\n",
    "test_df = normalize_test_features(test_df, scalers=scalers)\n",
    "test_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4e3933c7",
   "metadata": {},
   "source": [
    "### Prepare data for time series supervised learning (function to create sliding window)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "d52cc93a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# prepare data for time series\n",
    "\n",
    "# convert series to supervised learning\n",
    "def series_to_supervised(data, n_in=1, n_out=1, dropnan=True):\n",
    "    n_vars = 1 if type(data) is list else data.shape[1]\n",
    "    df = DataFrame(data)\n",
    "    cols, names = list(), list()\n",
    "    # input sequence (t-n, ... t-1)\n",
    "    for i in range(n_in, 0, -1):\n",
    "        cols.append(df.shift(i))\n",
    "        names += [('var%d(t-%d)' % (j+1, i)) for j in range(n_vars)]\n",
    "    # forecast sequence (t, t+1, ... t+n)\n",
    "    for i in range(0, n_out):\n",
    "        cols.append(df.shift(-i))\n",
    "        if i == 0:\n",
    "            names += [('var%d(t)' % (j+1)) for j in range(n_vars)]\n",
    "        else:\n",
    "            names += [('var%d(t+%d)' % (j+1, i)) for j in range(n_vars)]\n",
    "    # put it all together\n",
    "    agg = concat(cols, axis=1)\n",
    "    agg.columns = names\n",
    "    # drop rows with NaN values\n",
    "    if dropnan:\n",
    "        agg.dropna(inplace=True)\n",
    "    return agg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "87984320",
   "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>var1(t-10)</th>\n",
       "      <th>var2(t-10)</th>\n",
       "      <th>var3(t-10)</th>\n",
       "      <th>var4(t-10)</th>\n",
       "      <th>var5(t-10)</th>\n",
       "      <th>var6(t-10)</th>\n",
       "      <th>var7(t-10)</th>\n",
       "      <th>var8(t-10)</th>\n",
       "      <th>var9(t-10)</th>\n",
       "      <th>var10(t-10)</th>\n",
       "      <th>...</th>\n",
       "      <th>var1016(t)</th>\n",
       "      <th>var1017(t)</th>\n",
       "      <th>var1018(t)</th>\n",
       "      <th>var1019(t)</th>\n",
       "      <th>var1020(t)</th>\n",
       "      <th>var1021(t)</th>\n",
       "      <th>var1022(t)</th>\n",
       "      <th>var1023(t)</th>\n",
       "      <th>var1024(t)</th>\n",
       "      <th>var1025(t)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>201602</th>\n",
       "      <td>0.153002</td>\n",
       "      <td>-0.545807</td>\n",
       "      <td>-0.370448</td>\n",
       "      <td>0.304693</td>\n",
       "      <td>-0.024118</td>\n",
       "      <td>0.540703</td>\n",
       "      <td>0.174470</td>\n",
       "      <td>-0.667477</td>\n",
       "      <td>0.235845</td>\n",
       "      <td>0.535753</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000151</td>\n",
       "      <td>0.070641</td>\n",
       "      <td>-0.122372</td>\n",
       "      <td>-0.760683</td>\n",
       "      <td>-0.018953</td>\n",
       "      <td>-0.522950</td>\n",
       "      <td>-0.092673</td>\n",
       "      <td>0.164985</td>\n",
       "      <td>0.116053</td>\n",
       "      <td>-0.100273</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201603</th>\n",
       "      <td>0.311409</td>\n",
       "      <td>-0.269567</td>\n",
       "      <td>0.151433</td>\n",
       "      <td>-0.300232</td>\n",
       "      <td>-0.319257</td>\n",
       "      <td>0.187638</td>\n",
       "      <td>0.025189</td>\n",
       "      <td>0.388860</td>\n",
       "      <td>0.036709</td>\n",
       "      <td>-0.028918</td>\n",
       "      <td>...</td>\n",
       "      <td>0.126434</td>\n",
       "      <td>0.181488</td>\n",
       "      <td>-0.198964</td>\n",
       "      <td>-0.020205</td>\n",
       "      <td>0.130970</td>\n",
       "      <td>0.252293</td>\n",
       "      <td>-0.728723</td>\n",
       "      <td>0.270905</td>\n",
       "      <td>-0.716005</td>\n",
       "      <td>-0.176393</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201604</th>\n",
       "      <td>-0.519102</td>\n",
       "      <td>-0.340271</td>\n",
       "      <td>0.146133</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>-0.196062</td>\n",
       "      <td>-0.285319</td>\n",
       "      <td>0.733040</td>\n",
       "      <td>-0.109127</td>\n",
       "      <td>0.527685</td>\n",
       "      <td>0.219675</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.375093</td>\n",
       "      <td>-0.338485</td>\n",
       "      <td>0.509073</td>\n",
       "      <td>-0.501989</td>\n",
       "      <td>0.615657</td>\n",
       "      <td>0.648789</td>\n",
       "      <td>-0.508657</td>\n",
       "      <td>0.017548</td>\n",
       "      <td>-0.717804</td>\n",
       "      <td>-0.141793</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201605</th>\n",
       "      <td>0.929623</td>\n",
       "      <td>0.339709</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>-0.378638</td>\n",
       "      <td>-0.015372</td>\n",
       "      <td>0.090626</td>\n",
       "      <td>0.014490</td>\n",
       "      <td>-0.170956</td>\n",
       "      <td>0.555255</td>\n",
       "      <td>-0.006000</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.651840</td>\n",
       "      <td>-0.262113</td>\n",
       "      <td>-0.096112</td>\n",
       "      <td>-0.216066</td>\n",
       "      <td>-0.146001</td>\n",
       "      <td>0.146645</td>\n",
       "      <td>-0.900364</td>\n",
       "      <td>0.018970</td>\n",
       "      <td>0.224596</td>\n",
       "      <td>-0.100273</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201606</th>\n",
       "      <td>-0.243318</td>\n",
       "      <td>0.061918</td>\n",
       "      <td>0.300152</td>\n",
       "      <td>0.559433</td>\n",
       "      <td>0.258833</td>\n",
       "      <td>-0.808519</td>\n",
       "      <td>0.098995</td>\n",
       "      <td>0.062815</td>\n",
       "      <td>0.062930</td>\n",
       "      <td>-0.422597</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.128731</td>\n",
       "      <td>0.043607</td>\n",
       "      <td>-0.237798</td>\n",
       "      <td>-0.319142</td>\n",
       "      <td>0.213326</td>\n",
       "      <td>-0.296861</td>\n",
       "      <td>0.028832</td>\n",
       "      <td>0.481854</td>\n",
       "      <td>-0.222853</td>\n",
       "      <td>-0.127953</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\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",
       "      <td>...</td>\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",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201815</th>\n",
       "      <td>0.797333</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>-0.613434</td>\n",
       "      <td>-0.175824</td>\n",
       "      <td>-0.278700</td>\n",
       "      <td>0.119458</td>\n",
       "      <td>0.239677</td>\n",
       "      <td>0.346608</td>\n",
       "      <td>-0.475763</td>\n",
       "      <td>0.035077</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.535913</td>\n",
       "      <td>-0.480070</td>\n",
       "      <td>0.513649</td>\n",
       "      <td>0.189036</td>\n",
       "      <td>-0.151776</td>\n",
       "      <td>0.143130</td>\n",
       "      <td>-0.357305</td>\n",
       "      <td>-0.639151</td>\n",
       "      <td>0.477906</td>\n",
       "      <td>-0.972396</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201816</th>\n",
       "      <td>0.161418</td>\n",
       "      <td>0.059350</td>\n",
       "      <td>-0.156891</td>\n",
       "      <td>-0.213013</td>\n",
       "      <td>0.280339</td>\n",
       "      <td>0.312166</td>\n",
       "      <td>0.115440</td>\n",
       "      <td>-0.833810</td>\n",
       "      <td>-0.187396</td>\n",
       "      <td>-0.351217</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.475471</td>\n",
       "      <td>0.236124</td>\n",
       "      <td>0.353747</td>\n",
       "      <td>0.397505</td>\n",
       "      <td>0.451361</td>\n",
       "      <td>-0.138578</td>\n",
       "      <td>-0.306785</td>\n",
       "      <td>0.371796</td>\n",
       "      <td>-0.192488</td>\n",
       "      <td>-0.972396</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201817</th>\n",
       "      <td>0.887821</td>\n",
       "      <td>-0.180743</td>\n",
       "      <td>-0.155074</td>\n",
       "      <td>-0.018393</td>\n",
       "      <td>-0.760965</td>\n",
       "      <td>0.572115</td>\n",
       "      <td>-0.546911</td>\n",
       "      <td>-0.084733</td>\n",
       "      <td>0.133753</td>\n",
       "      <td>-0.301270</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.073411</td>\n",
       "      <td>0.135366</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>-0.207863</td>\n",
       "      <td>0.403494</td>\n",
       "      <td>0.488648</td>\n",
       "      <td>-0.299530</td>\n",
       "      <td>-0.063334</td>\n",
       "      <td>0.803492</td>\n",
       "      <td>-0.979297</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201818</th>\n",
       "      <td>0.554046</td>\n",
       "      <td>0.137046</td>\n",
       "      <td>0.247404</td>\n",
       "      <td>0.092975</td>\n",
       "      <td>-0.421474</td>\n",
       "      <td>-0.790012</td>\n",
       "      <td>-0.537559</td>\n",
       "      <td>0.433995</td>\n",
       "      <td>0.491965</td>\n",
       "      <td>0.484942</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.024630</td>\n",
       "      <td>-0.198176</td>\n",
       "      <td>-0.160777</td>\n",
       "      <td>-0.014544</td>\n",
       "      <td>0.017891</td>\n",
       "      <td>0.102530</td>\n",
       "      <td>0.360620</td>\n",
       "      <td>-0.128165</td>\n",
       "      <td>-0.316256</td>\n",
       "      <td>-0.965494</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>201819</th>\n",
       "      <td>0.149763</td>\n",
       "      <td>-0.577400</td>\n",
       "      <td>-0.084736</td>\n",
       "      <td>-0.286580</td>\n",
       "      <td>-0.085460</td>\n",
       "      <td>-0.189519</td>\n",
       "      <td>-0.881927</td>\n",
       "      <td>0.317386</td>\n",
       "      <td>-0.189051</td>\n",
       "      <td>0.544778</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.190107</td>\n",
       "      <td>-0.052215</td>\n",
       "      <td>0.255767</td>\n",
       "      <td>-0.741897</td>\n",
       "      <td>0.086524</td>\n",
       "      <td>0.102446</td>\n",
       "      <td>0.620181</td>\n",
       "      <td>0.121715</td>\n",
       "      <td>0.747221</td>\n",
       "      <td>-0.979297</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>122 rows × 11275 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        var1(t-10)  var2(t-10)  var3(t-10)  var4(t-10)  var5(t-10)  \\\n",
       "201602    0.153002   -0.545807   -0.370448    0.304693   -0.024118   \n",
       "201603    0.311409   -0.269567    0.151433   -0.300232   -0.319257   \n",
       "201604   -0.519102   -0.340271    0.146133   -1.000000   -0.196062   \n",
       "201605    0.929623    0.339709   -1.000000   -0.378638   -0.015372   \n",
       "201606   -0.243318    0.061918    0.300152    0.559433    0.258833   \n",
       "...            ...         ...         ...         ...         ...   \n",
       "201815    0.797333   -1.000000   -0.613434   -0.175824   -0.278700   \n",
       "201816    0.161418    0.059350   -0.156891   -0.213013    0.280339   \n",
       "201817    0.887821   -0.180743   -0.155074   -0.018393   -0.760965   \n",
       "201818    0.554046    0.137046    0.247404    0.092975   -0.421474   \n",
       "201819    0.149763   -0.577400   -0.084736   -0.286580   -0.085460   \n",
       "\n",
       "        var6(t-10)  var7(t-10)  var8(t-10)  var9(t-10)  var10(t-10)  ...  \\\n",
       "201602    0.540703    0.174470   -0.667477    0.235845     0.535753  ...   \n",
       "201603    0.187638    0.025189    0.388860    0.036709    -0.028918  ...   \n",
       "201604   -0.285319    0.733040   -0.109127    0.527685     0.219675  ...   \n",
       "201605    0.090626    0.014490   -0.170956    0.555255    -0.006000  ...   \n",
       "201606   -0.808519    0.098995    0.062815    0.062930    -0.422597  ...   \n",
       "...            ...         ...         ...         ...          ...  ...   \n",
       "201815    0.119458    0.239677    0.346608   -0.475763     0.035077  ...   \n",
       "201816    0.312166    0.115440   -0.833810   -0.187396    -0.351217  ...   \n",
       "201817    0.572115   -0.546911   -0.084733    0.133753    -0.301270  ...   \n",
       "201818   -0.790012   -0.537559    0.433995    0.491965     0.484942  ...   \n",
       "201819   -0.189519   -0.881927    0.317386   -0.189051     0.544778  ...   \n",
       "\n",
       "        var1016(t)  var1017(t)  var1018(t)  var1019(t)  var1020(t)  \\\n",
       "201602    0.000151    0.070641   -0.122372   -0.760683   -0.018953   \n",
       "201603    0.126434    0.181488   -0.198964   -0.020205    0.130970   \n",
       "201604   -0.375093   -0.338485    0.509073   -0.501989    0.615657   \n",
       "201605   -0.651840   -0.262113   -0.096112   -0.216066   -0.146001   \n",
       "201606   -0.128731    0.043607   -0.237798   -0.319142    0.213326   \n",
       "...            ...         ...         ...         ...         ...   \n",
       "201815   -0.535913   -0.480070    0.513649    0.189036   -0.151776   \n",
       "201816   -0.475471    0.236124    0.353747    0.397505    0.451361   \n",
       "201817   -0.073411    0.135366   -1.000000   -0.207863    0.403494   \n",
       "201818   -0.024630   -0.198176   -0.160777   -0.014544    0.017891   \n",
       "201819   -0.190107   -0.052215    0.255767   -0.741897    0.086524   \n",
       "\n",
       "        var1021(t)  var1022(t)  var1023(t)  var1024(t)  var1025(t)  \n",
       "201602   -0.522950   -0.092673    0.164985    0.116053   -0.100273  \n",
       "201603    0.252293   -0.728723    0.270905   -0.716005   -0.176393  \n",
       "201604    0.648789   -0.508657    0.017548   -0.717804   -0.141793  \n",
       "201605    0.146645   -0.900364    0.018970    0.224596   -0.100273  \n",
       "201606   -0.296861    0.028832    0.481854   -0.222853   -0.127953  \n",
       "...            ...         ...         ...         ...         ...  \n",
       "201815    0.143130   -0.357305   -0.639151    0.477906   -0.972396  \n",
       "201816   -0.138578   -0.306785    0.371796   -0.192488   -0.972396  \n",
       "201817    0.488648   -0.299530   -0.063334    0.803492   -0.979297  \n",
       "201818    0.102530    0.360620   -0.128165   -0.316256   -0.965494  \n",
       "201819    0.102446    0.620181    0.121715    0.747221   -0.979297  \n",
       "\n",
       "[122 rows x 11275 columns]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# length of window\n",
    "days = 10\n",
    "\n",
    "# frame as supervised learning\n",
    "train = series_to_supervised(train_df, n_in=days)\n",
    "test = series_to_supervised(test_df, n_in=days)\n",
    "\n",
    "DataFrame(train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59dce035",
   "metadata": {},
   "source": [
    "### Features and Labels Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "548d8f4f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def features_labels_set(timeseries_data, original_df):\n",
    "    \n",
    "    \"\"\" Features \"\"\"\n",
    "    # We define the number of features as (Cases and media cloud)\n",
    "    n_features = original_df.shape[1]\n",
    "\n",
    "    # The features to train the model will be all except the values of the actual week \n",
    "    # We can't use other variables in week t because whe need to resample a a 3D Array\n",
    "    features_set = DataFrame(timeseries_data.values[:,:-n_features])\n",
    "    # Convert pandas data frame to np.array to reshape as 3D Array\n",
    "    features_set = features_set.to_numpy()\n",
    "    print(f'The shape of the features is {features_set.shape}')\n",
    "    \n",
    "    \"\"\" Labels \"\"\"\n",
    "    # We will use Covid cases in last week \n",
    "    labels_set = DataFrame(timeseries_data.values[:,-1])\n",
    "    # Convert pandas data frame to np.array\n",
    "    labels_set = labels_set.to_numpy()\n",
    "    print(f'The shape of the labels is {labels_set.shape}')\n",
    "    \n",
    "    return features_set, labels_set, n_features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "a97fc870",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train:\n",
      "The shape of the features is (122, 10250)\n",
      "The shape of the labels is (122, 1)\n",
      "Test:\n",
      "The shape of the features is (23, 10250)\n",
      "The shape of the labels is (23, 1)\n"
     ]
    }
   ],
   "source": [
    "# Train features and labels set\n",
    "print('Train:')\n",
    "train_X, train_y, n_features = features_labels_set(timeseries_data=train, original_df=dengue_df)\n",
    "\n",
    "# Test features and labels set\n",
    "print('Test:')\n",
    "test_X, test_y, n_features = features_labels_set(timeseries_data=test, original_df=dengue_df)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6ed1b8d4",
   "metadata": {},
   "source": [
    "# Modeling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "3225806c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def reshape_tensor(train_X, test_X, n_features):\n",
    "    print('The initial shapes are:')\n",
    "    print(f'The train shape is {train_X.shape}')\n",
    "    print(f'The test shape is {test_X.shape}')\n",
    "    \n",
    "    # reshape input to be 3D [samples, timesteps, features]\n",
    "    train_X = train_X.reshape((train_X.shape[0], days, n_features))\n",
    "    test_X = test_X.reshape((test_X.shape[0], days, n_features))\n",
    "    \n",
    "    print('-----------------------')\n",
    "    print('The Final shapes are:')\n",
    "    print(f'The train shape is {train_X.shape}')\n",
    "    print(f'The test shape is {test_X.shape}')\n",
    "    \n",
    "    return train_X, test_X"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "4ce5ca01",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The initial shapes are:\n",
      "The train shape is (122, 10250)\n",
      "The test shape is (23, 10250)\n",
      "-----------------------\n",
      "The Final shapes are:\n",
      "The train shape is (122, 10, 1025)\n",
      "The test shape is (23, 10, 1025)\n"
     ]
    }
   ],
   "source": [
    "# reshape input to be 3D [samples, timesteps, features]\n",
    "train_X, test_X = reshape_tensor(train_X, test_X, n_features)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "92094dcb",
   "metadata": {},
   "source": [
    "# Define the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "3414d34c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set Seed\n",
    "tf.random.set_seed(0)\n",
    "\n",
    "# design network\n",
    "model = Sequential()\n",
    "model.add(LSTM(120, dropout=0.1, input_shape=(train_X.shape[1], train_X.shape[2]), return_sequences=True))\n",
    "model.add(LSTM(240, dropout=0.1, input_shape=(train_X.shape[1], 120)))\n",
    "model.add(Dense(60))\n",
    "model.add(Dense(1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "ce51388b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm (LSTM)                  (None, 10, 120)           550080    \n",
      "_________________________________________________________________\n",
      "lstm_1 (LSTM)                (None, 240)               346560    \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 60)                14460     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 61        \n",
      "=================================================================\n",
      "Total params: 911,161\n",
      "Trainable params: 911,161\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "opt = keras.optimizers.Adam()\n",
    "#metric = tf.keras.metrics.AUC(name='auc', multi_label=True, num_labels=3)\n",
    "\n",
    "# Compile the model\n",
    "model.compile(loss='categorical_crossentropy', optimizer=opt, metrics='mae')\n",
    "\n",
    "# Print the model\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "72dd5e74",
   "metadata": {},
   "source": [
    "### Train the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "c5db5cbd",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "\n",
    "# EarlyStopping:\n",
    "monitor = EarlyStopping(monitor='val_loss', min_delta=1e-3, patience=20, \n",
    "        verbose=1, mode='auto', restore_best_weights=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "58962f5d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "2/2 - 2s - loss: -5.8667e-08 - mae: 0.7304 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 2/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7325 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 3/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7303 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 4/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7261 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 5/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7249 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 6/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7239 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 7/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7289 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 8/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7286 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 9/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7303 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 10/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7263 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 11/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7293 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 12/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7271 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 13/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7260 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 14/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7265 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 15/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7273 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 16/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7320 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 17/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7272 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 18/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7309 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 19/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7262 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 20/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7282 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Epoch 21/50\n",
      "2/2 - 0s - loss: -5.8667e-08 - mae: 0.7287 - val_loss: -1.1595e-07 - val_mae: 0.9700\n",
      "Restoring model weights from the end of the best epoch.\n",
      "Epoch 00021: early stopping\n"
     ]
    }
   ],
   "source": [
    "# fit network\n",
    "history = model.fit(train_X, train_y, epochs=50, batch_size=64, validation_data=(test_X, test_y), verbose=2, shuffle=False, callbacks=[monitor])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "1d59797e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot history\n",
    "plt.plot(history.history['loss'], label='train')\n",
    "plt.plot(history.history['val_loss'], label='validation')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ce4e745",
   "metadata": {},
   "source": [
    "# Test the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "e97c5e2f",
   "metadata": {},
   "outputs": [],
   "source": [
    "#print(f'The scalers are: {scalers.keys()}')\n",
    "y_scaler = scalers['scaler_Labels']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "9caa18c5",
   "metadata": {},
   "outputs": [],
   "source": [
    "from math import sqrt\n",
    "from numpy import concatenate\n",
    "\n",
    "def test_model(model, test_X, test_y, scaler, rnn = None):\n",
    "    \n",
    "    # If model is a classical machine learning model and test_X is a 3D tensor, then convert to 2D\n",
    "    if not rnn and (len(test_X.shape) == 3):\n",
    "        test_X = test_X.reshape((test_X.shape[0], -1))\n",
    "    \n",
    "    # do the prediction\n",
    "    yhat = model.predict(test_X)\n",
    "    \n",
    "    # Invert scaling for forecast\n",
    "    # Inverse Scaler\n",
    "    \n",
    "    # Predicted\n",
    "    if not rnn:\n",
    "        yhat = yhat.reshape(-1, 1)\n",
    "    \n",
    "    inv_yhat = scaler.inverse_transform(yhat)\n",
    "    \n",
    "    # Real:\n",
    "    inv_y = scaler.inverse_transform(test_y)\n",
    "    \n",
    "    return inv_yhat, inv_y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "c19d485d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# With LSTM:\n",
    "inv_yhat_lstm, inv_y_lstm = test_model(model=model, test_X=test_X, test_y=test_y, scaler=y_scaler, rnn = True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "872f536a",
   "metadata": {},
   "source": [
    "### Mean Absolute Percentage Error (MAPE)\n",
    "\n",
    "$$\n",
    "MAPE = \\displaystyle\\frac{100\\%}{n}\\sum_{t=1}^{n}\\left |\\frac{x_i-y_i}{y_t}\\right|\n",
    "$$\n",
    "\n",
    "MAPE has a problem if there are zeros in the test data, so other metrics can be explored"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "b26f81f6",
   "metadata": {},
   "outputs": [],
   "source": [
    "def mean_absolute_percentage_error(y_true, y_pred):\n",
    "    y_true, y_pred = np.array(y_true), np.array(y_pred)\n",
    "    mape = np.mean(np.abs((y_true - y_pred) / y_true)) * 100\n",
    "    print('Test MAPE: %.3f' % mape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "f0ba328c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--- LSTM ---\n",
      "Test MAPE: 1387.426\n"
     ]
    }
   ],
   "source": [
    "# calculate MAPE\n",
    "print('--- LSTM ---')\n",
    "mean_absolute_percentage_error(inv_y_lstm, inv_yhat_lstm)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dd64e61b",
   "metadata": {},
   "source": [
    "### Symmetric Mean Absolute Percentage Error (sMAPE)\n",
    "\n",
    "$$\n",
    "sMAPE = \\displaystyle\\frac{100\\%}{n}\\sum_{t=1}^{n} \\frac{|x_i-y_i|}{|x_i|+|y_t|}\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "4bc9188e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def symmetric_mean_absolute_percentage_error(y_true, y_pred):\n",
    "    \n",
    "    y_true, y_pred = np.array(y_true), np.array(y_pred)\n",
    "    smape = 1/len(y_true) * np.sum(2 * np.abs(y_pred-y_true) / (np.abs(y_true) + np.abs(y_pred))*100)\n",
    "    print('Test sMAPE: %.3f' % smape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "31610c6b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--- LSTM ---\n",
      "Test sMAPE: 173.037\n"
     ]
    }
   ],
   "source": [
    "print('--- LSTM ---')\n",
    "symmetric_mean_absolute_percentage_error(inv_y_lstm, inv_yhat_lstm)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bf93ad52",
   "metadata": {},
   "source": [
    "### Mean Absoulte Error (MAE)\n",
    "$$\n",
    "MAE = \\sum_{i=1}^{D}|x_i-y_i|\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "3e05a3a0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--- LSTM ---\n",
      "Test MAE: 0.000\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import  mean_absolute_error\n",
    "\n",
    "# calculate MAE\n",
    "print('--- LSTM ---')\n",
    "mae = mean_absolute_error(inv_y_lstm, inv_yhat_lstm)\n",
    "print('Test MAE: %.3f' % mae)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "f80356f0",
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_predictions(inv_y, inv_yhat, model_name = ''):\n",
    "    data_predict = inv_yhat  ## predicted target  cases\n",
    "    dataY_plot = inv_y  ##  real test-target cases\n",
    "\n",
    "    data_predict = data_predict.reshape(len(data_predict), 1)\n",
    "    dataY_plot = dataY_plot.reshape(len(dataY_plot), 1)\n",
    "\n",
    "    plt.plot(dataY_plot, label = 'actual')\n",
    "    plt.plot(data_predict, label = 'predicted')\n",
    "    plt.legend(loc=\"upper left\")\n",
    "\n",
    "    plt.suptitle(f'Time-Series Prediction with {model_name}')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "7a8f3449",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkYAAAHNCAYAAAD7QrTqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAA9hAAAPYQGoP6dpAABu8klEQVR4nO3deXwTZf4H8E/SNklb2qSlR1ootNw3RZBaOZVqUUTqsgtWRGCrRQURWVdhFQqrKwjouiiH6AquP+TyAEFAawG5SsFyU0SOImfaQknS+0ie3x8zHRpaoMGWlvbzfr3mFTLzncwzCSEfnnlmRiWEECAiIiIiqGu7AURERER1BYMRERERkYzBiIiIiEjGYEREREQkYzAiIiIikjEYEREREckYjIiIiIhkDEZEREREMgYjIiIiIhmDETVIo0ePRmhoaG03o1apVCpMnz69tptxR1X2uVf3+9C/f3/079+/2l6vpjnzXRg9ejQaNWpUsw0iqmUMRlRvqFSqKk1bt26t7abe0Lp169CvXz8EBATAw8MDLVq0wLBhw7Bp06babtofdubMGYfPwcXFBc2aNcMTTzyBAwcO1HbznJKWlobp06fjzJkztd2Uapefn4/p06fXyPekf//+6NSp0y3rduzYgUceeQRNmjSBTqdDs2bNMHjwYHz55ZcApIBWle/66NGjle2qVCq0bt260u0lJiYq63z11VfVtr90d3Kt7QYQVZcvvvjC4fn//vc/JCYmVpjfvn17fPLJJ7Db7Xeyebc0d+5c/P3vf0e/fv0wZcoUeHh44OTJk/jpp5+wYsUKDBw4sFq3V1BQAFfXO/9PQGxsLB599FHYbDYcO3YMCxcuxMaNG7F7926Eh4ff8fbczvuQlpaGGTNmoH///hV6W3788cdqbF3Nu/67kJ+fjxkzZgBArfR8rV69GsOHD0d4eDhefvll+Pj4ID09Hdu2bcMnn3yCp556CmPHjkVUVJSyTnp6OqZNm4b4+Hj06dNHmd+yZUvlzzqdDidPnsSePXvQs2dPh20uW7YMOp0OhYWFNb+DVOcxGFG98fTTTzs83717NxITEyvMr4tKS0vx1ltv4aGHHqr0hzUzM7NatmO321FcXAydTgedTlctr+mse+65x+Ez6dWrFx5//HEsXLgQH3/8caXr5OXlwdPTs0baU93vg0ajqdbXq2lubm613QQH06dPR4cOHbB79+4K72XZ9yAyMhKRkZHK/F9++QXTpk1DZGTkDb/vLVu2RGlpKZYvX+4QjAoLC/Htt99i0KBB+Prrr2tgj+huw0Np1CBdP66i7DDP3LlzMX/+fLRo0QIeHh54+OGHce7cOQgh8NZbb6Fp06Zwd3fHkCFDkJ2dXeF1N27ciD59+sDT0xNeXl4YNGgQjh49esv2XL58GVarFb169ap0eUBAgMPzoqIiJCQkoFWrVtBqtQgJCcFrr72GoqIihzqVSoXx48dj2bJl6NixI7RarXJYrrKxNRcuXMBf//pXBAYGQqvVomPHjvjss88qtOfDDz9Ex44d4eHhAR8fH/To0UM5zOGsBx98EID0v34AWLp0KVQqFX7++We8+OKLCAgIQNOmTZX6qr7Ha9asQadOnaDT6dCpUyd8++23lW7/Ru9DXFwcgoODodVqERYWhhdeeAHFxcVYunQp/vKXvwAAHnjggQqHaCsbY5SZmYm4uDgEBgZCp9Oha9eu+Pzzzx1qyv8dXLx4MVq2bAmtVot7770Xe/fuvel7aDab4eLignnz5inzLl++DLVajcaNG0MIocx/4YUXYDQaleflvwtnzpyBv78/AGDGjBnKvlX2/sTExKBRo0bw9/fHq6++CpvNdtM2VtWpU6dw7733Vhowr/8eOCs2NhYrV6506CFbt24d8vPzMWzYsD/02lR/sMeIqJxly5ahuLgYL730ErKzszF79mwMGzYMDz74ILZu3YrXX38dJ0+exIcffohXX33VITR88cUXGDVqFKKjo/Huu+8iPz8fCxcuRO/evbF///6bDnANCAiAu7s71q1bh5deegm+vr43rLXb7Xj88cexY8cOxMfHo3379jh8+DD+/e9/47fffsOaNWsc6jdv3oxVq1Zh/Pjx8PPzu2E7MjIycN999ylhyt/fHxs3bkRcXBysVismTpwIQDr0MmHCBPz5z3/Gyy+/jMLCQhw6dAgpKSl46qmnqvpWK06dOgUAaNy4scP8F198Ef7+/pg2bRry8vIAVP09/vHHHzF06FB06NABM2fOxJUrVzBmzBiHgHUjFy9eRM+ePWE2mxEfH4927drhwoUL+Oqrr5Cfn4++fftiwoQJmDdvHv7xj3+gffv2AKA8Xq+goAD9+/fHyZMnMX78eISFhWH16tUYPXo0zGYzXn75ZYf6L7/8Ejk5ORg7dixUKhVmz56NP/3pTzh9+vQNe3cMBgM6deqEbdu2YcKECQCkcToqlQrZ2dlIS0tDx44dAQDbt293ONxUnr+/PxYuXIgXXngBTzzxBP70pz8BALp06aLU2Gw2REdHIyIiAnPnzsVPP/2E9957Dy1btsQLL7xwy/f3Vpo3b46kpCScP3++Sp+XM5566ill/FRZIP/yyy8xYMCAPxy6qB4RRPXUuHHjxI3+io8aNUo0b95ceZ6eni4ACH9/f2E2m5X5U6ZMEQBE165dRUlJiTI/NjZWaDQaUVhYKIQQIicnRxgMBvHcc885bMdkMgm9Xl9hfmWmTZsmAAhPT0/xyCOPiH/9618iNTW1Qt0XX3wh1Gq12L59u8P8RYsWCQBi586dyjwAQq1Wi6NHj1Z4HQAiISFBeR4XFyeCgoLE5cuXHeqefPJJodfrRX5+vhBCiCFDhoiOHTvecn+uV/Yez5gxQ2RlZQmTySS2bt0qunXrJgCIr7/+WgghxJIlSwQA0bt3b1FaWqqs78x7HB4eLoKCghw+yx9//FEAcPjcK3sfnnnmGaFWq8XevXsr7IPdbhdCCLF69WoBQGzZsqVCTb9+/US/fv2U5x988IEAIP7v//5PmVdcXCwiIyNFo0aNhNVqdXh/GjduLLKzs5XatWvXCgBi3bp1FbZV3rhx40RgYKDyfNKkSaJv374iICBALFy4UAghxJUrV4RKpRL/+c9/lLrrvwtZWVkV3pPytQDEP//5T4f53bp1E927d79p+4SQ3ptb/d3573//KwAIjUYjHnjgATF16lSxfft2YbPZbrjO3r17BQCxZMmSW263R48eIi4uTgghxNWrV4VGoxGff/652LJliwAgVq9efcv9oPqNh9KIyvnLX/4CvV6vPI+IiAAgjV8qP0A3IiICxcXFuHDhAgDprBaz2YzY2FhcvnxZmVxcXBAREYEtW7bcctszZszAl19+iW7duuGHH37AG2+8ge7du+Oee+7BsWPHlLrVq1ejffv2aNeuncO2yv4HfP22+vXrhw4dOtx020IIfP311xg8eDCEEA6vGx0dDYvFgn379gGQeifOnz9/y8M7N5KQkAB/f38YjUb0798fp06dwrvvvqv0TpR57rnn4OLiojyv6nt86dIlHDhwAKNGjXL4LB966KFbvg92ux1r1qzB4MGD0aNHjwrLVSqV0/u7YcMGGI1GxMbGKvPc3NwwYcIE5Obm4ueff3aoHz58OHx8fJTnZb07p0+fvul2+vTpg4yMDBw/fhyA1DPUt29f9OnTB9u3bwcg9SIJIW7YY1RVzz//fIVt36p9VfXXv/4VmzZtQv/+/bFjxw689dZb6NOnD1q3bo1du3b94dd/6qmn8M0336C4uBhfffUVXFxc8MQTT1RDy6m+4KE0onKaNWvm8LzshzUkJKTS+VevXgUAnDhxAsC18TLX8/b2BiAdVrFYLA7Lyo/3iI2NRWxsLKxWK1JSUrB06VJ8+eWXGDx4MI4cOQKdTocTJ07g2LFjyliQ610/UDssLOzGOyzLysqC2WzG4sWLsXjx4pu+7uuvv46ffvoJPXv2RKtWrfDwww/jqaeeuuH4qOvFx8fjL3/5C9RqNQwGgzL26XrXt7uq7/Hvv/8OAJWemt22bVsl4FUmKysLVqu1SqeUV9Xvv/+O1q1bQ612/H9o2aG3svaWuf7vYFlIKvu7diNlYWf79u1o2rQp9u/fj7fffhv+/v6YO3eusszb2xtdu3a97f3R6XQV/u75+Pjcsn3OiI6ORnR0NPLz85GamoqVK1di0aJFeOyxx/Drr7/+ocNeTz75JF599VVs3LgRy5Ytw2OPPQYvL69qazvd/RiMiMop30NRlflCHtRaNpjziy++cAg6Zcp6m1auXIkxY8ZU+hrleXt746GHHsJDDz0ENzc3fP7550hJSUG/fv1gt9vRuXNnvP/++5W26foQ5+7uXmldeWXtf/rppzFq1KhKa8rGmbRv3x7Hjx/H+vXrsWnTJnz99ddYsGABpk2bppzmfTOtW7d2ONX6Rq5vd1Xf47vdrf6u3UhwcDDCwsKwbds2hIaGQgiByMhI+Pv74+WXX8bvv/+O7du34/77768Q0qqjfTXBw8MDffr0QZ8+feDn54cZM2Zg48aNN/w7WhVBQUHo378/3nvvPezcuZNnolEF9eNfEqJaVna9lICAgJv+6EdHRyMxMdGp1+7Rowc+//xzXLp0SdnWwYMHMWDAgNs6tFMZf39/eHl5wWazVSm0eHp6Yvjw4Rg+fDiKi4vxpz/9Cf/6178wZcqUGrsMQFXf4+bNmwO41sNUXtlhphvx9/eHt7c3jhw5ctM6Z9735s2b49ChQ7Db7Q6B5Ndff3Vob3Xo06cPtm3bhrCwMISHh8PLywtdu3aFXq/Hpk2bsG/fvluG1+r6O1Xdyg5tln0P/oinnnoKzz77LAwGAx599NE//HpUv3CMEVE1iI6Ohre3N9555x2UlJRUWJ6VlQVA+t9qVFSUwwRIF9VLTk6u9LU3btwIQDoMBADDhg3DhQsX8Mknn1SoLSgoUM7gcoaLiwuGDh2Kr7/+utJQUNZ+ALhy5YrDMo1Ggw4dOkAIUem+Vxdn3uPw8HB8/vnnDoctExMTkZaWdtNtqNVqxMTEYN26dfjll18qLC/rtSm7ppLZbL5lux999FGYTCasXLlSmVdaWooPP/wQjRo1Qr9+/W75GlXVp08fnDlzBitXrlQOranVatx///14//33UVJScsvxRR4eHgCqtm81ISkpqdL5GzZsAHDte/BH/PnPf0ZCQgIWLFhw1113imoee4yIqoG3tzcWLlyIkSNH4p577sGTTz4Jf39/nD17Ft9//z169eqFjz766Ibr5+fn4/7778d9992HgQMHIiQkBGazGWvWrMH27dsRExODbt26AQBGjhyJVatW4fnnn8eWLVvQq1cv2Gw2/Prrr1i1ahV++OGHSgcO38qsWbOwZcsWRERE4LnnnkOHDh2QnZ2Nffv24aefflKu2/Twww/DaDSiV69eCAwMxLFjx/DRRx9h0KBBNTpWw5n3eObMmRg0aBB69+6Nv/71r8jOzlauvZSbm3vT7bzzzjv48ccf0a9fP+VyCJcuXcLq1auxY8cOGAwGhIeHw8XFBe+++y4sFgu0Wi0efPDBSse+xMfH4+OPP8bo0aORmpqK0NBQfPXVV9i5cyc++OCDan3PykLP8ePH8c477yjz+/bti40bNyrXRboZd3d3dOjQAStXrkSbNm3g6+uLTp06Vdu4q6ysLLz99tsV5oeFhWHEiBEYMmQIwsLCMHjwYLRs2RJ5eXn46aefsG7dOtx7770YPHjwH26DXq9vcPcJpKpjMCKqJk899RSCg4Mxa9YszJkzB0VFRWjSpAn69OlTYVzR9QwGAz755BN8//33WLJkCUwmE1xcXNC2bVvMmTNHuTYNIPUArFmzBv/+97/xv//9D99++61yX7WXX34Zbdq0ua32BwYGYs+ePfjnP/+Jb775BgsWLEDjxo3RsWNHvPvuu0rd2LFjsWzZMrz//vvIzc1F06ZNMWHCBLz55pu3tV1nVPU9HjhwIFavXo0333wTU6ZMQcuWLbFkyRKsXbv2lvcAa9KkCVJSUjB16lQsW7YMVqsVTZo0wSOPPKL0phiNRixatAgzZ85EXFwcbDYbtmzZUmkwcnd3x9atWzF58mR8/vnnsFqtaNu2LZYsWaLcy6u6tG3bFgEBAcjMzETv3r2V+WWBqWfPnpUOdL/ep59+ipdeegmvvPIKiouLkZCQUG3BKDMzE1OnTq0wf8CAARgxYgQ+/fRTrF27FqtWrcLFixchhECLFi3wxhtv4PXXX683Y8mo7lKJW43oIyIiImogOMaIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQkYzAiIiIikjEYEREREckYjIiIiIhkDEZEREREMgYjIiIiIhmDEREREZGMwYiIiIhIxmBEREREJGMwIiIiIpIxGBERERHJGIyIiIiIZAxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkcy1thtwt7Hb7bh48SK8vLygUqlquzlERERUBUII5OTkIDg4GGr1jfuFGIycdPHiRYSEhNR2M4iIiOg2nDt3Dk2bNr3hcgYjJ3l5eQGQ3lhvb+9abg0RERFVhdVqRUhIiPI7fiMMRk4qO3zm7e3NYERERHSXudUwGA6+JiIiIpIxGBERERHJGIyIiIiIZBxjVANsNhtKSkpquxn0B7i4uMDV1ZWXZCAiamAYjKpZbm4uzp8/DyFEbTeF/iAPDw8EBQVBo9HUdlOIiOgOYTCqRjabDefPn4eHhwf8/f3Z23CXEkKguLgYWVlZSE9PR+vWrW96MTAiIqo/GIyqUUlJCYQQ8Pf3h7u7e203h/4Ad3d3uLm54ffff0dxcTF0Ol1tN4mIiO4A/je4BrCnqH5gLxERUcPDf/mJiIiIZLcVjObPn4/Q0FDodDpERERgz549N61fvXo12rVrB51Oh86dO2PDhg0Oy4UQmDZtGoKCguDu7o6oqCicOHHCoSY7OxsjRoyAt7c3DAYD4uLikJubqywvLCzE6NGj0blzZ7i6uiImJuambdq5cydcXV0RHh7u1L7TnTV69OhbfpZERETVxelgtHLlSkyaNAkJCQnYt28funbtiujoaGRmZlZav2vXLsTGxiIuLg779+9HTEwMYmJicOTIEaVm9uzZmDdvHhYtWoSUlBR4enoiOjoahYWFSs2IESNw9OhRJCYmYv369di2bRvi4+OV5TabDe7u7pgwYQKioqJuug9msxnPPPMMBgwY4OzuUyWmT5/OgElERPWC08Ho/fffx3PPPYcxY8agQ4cOWLRoETw8PPDZZ59VWv+f//wHAwcOxN///ne0b98eb731Fu655x589NFHAKTeog8++ABvvvkmhgwZgi5duuB///sfLl68iDVr1gAAjh07hk2bNuHTTz9FREQEevfujQ8//BArVqzAxYsXAQCenp5YuHAhnnvuORiNxpvuw/PPP4+nnnoKkZGRzu4+ERER1WNOBaPi4mKkpqY69Mio1WpERUUhOTm50nWSk5Mr9OBER0cr9enp6TCZTA41er0eERERSk1ycjIMBgN69Oih1ERFRUGtViMlJcWZXcCSJUtw+vRpJCQkVKm+qKgIVqvVYaoRuVmA5QJQlAvUwjWQNm3ahN69e8NgMKBx48Z47LHHcOrUKWX5+fPnERsbC19fX3h6eqJHjx5ISUnB0qVLMWPGDBw8eBAqlQoqlQpLly7FmTNnoFKpcODAAeU1zGYzVCoVtm7dCkDq5YuLi0NYWBjc3d3Rtm1b/Oc//7nDe05ERHSNU6frX758GTabDYGBgQ7zAwMD8euvv1a6jslkqrTeZDIpy8vm3awmICDAseGurvD19VVqquLEiROYPHkytm/fDlfXqu36zJkzMWPGjCpvozwhBApKbFUrtmQCpYUATIDaFdB6Azo9oGkE3MbZUe5uLk6dHZeXl4dJkyahS5cuyM3NxbRp0/DEE0/gwIEDyM/PR79+/dCkSRN89913MBqN2LdvH+x2O4YPH44jR45g06ZN+OmnnwBIwTYjI+OW27Tb7WjatClWr16Nxo0bY9euXYiPj0dQUBCGDRvm9D4TERH9UQ3mOkY2mw1PPfUUZsyYgTZt2lR5vSlTpmDSpEnKc6vVipCQkCqtW1BiQ4dpPzjd1uqQ9s9oeGiq/vEOHTrU4flnn30Gf39/pKWlYdeuXcjKysLevXvh6+sLAGjVqpVS26hRI7i6ut7yEOb13NzcHEJnWFgYkpOTsWrVKgYjIiKqFU4FIz8/P7i4uFToDcjIyLjhj6LRaLxpfdljRkYGgoKCHGrKBvQajcYKg7tLS0uRnZ1d5R/jnJwc/PLLL9i/fz/Gjx8PQOqxEELA1dUVP/74Ix588MEK62m1Wmi12ipt42524sQJTJs2DSkpKbh8+TLsdjsA4OzZszhw4AC6deumhKLqNH/+fHz22Wc4e/YsCgoKUFxczIHcRERUa5wKRhqNBt27d0dSUpJyCrXdbkdSUpISNq4XGRmJpKQkTJw4UZmXmJioDHwOCwuD0WhEUlKS8oNotVqRkpKCF154QXkNs9mM1NRUdO/eHQCwefNm2O12REREVKnt3t7eOHz4sMO8BQsWYPPmzfjqq68QFhZW1behytzdXJD2z+jbfwEhgJJCoNACFFmB0gLH5a46QKuXDrm56YByh87c3Vyc2tTgwYPRvHlzfPLJJwgODobdbkenTp1QXFx8W1fxLrs4Yvl7xl1/Y90VK1bg1VdfxXvvvYfIyEh4eXlhzpw5To8bIyIiqi5OH0qbNGkSRo0ahR49eqBnz5744IMPkJeXhzFjxgAAnnnmGTRp0gQzZ84EALz88svo168f3nvvPQwaNAgrVqzAL7/8gsWLFwOQrhI9ceJEvP3222jdujXCwsIwdepUBAcHK+Grffv2GDhwIJ577jksWrQIJSUlGD9+PJ588kkEBwcrbUtLS0NxcTGys7ORk5OjDPwNDw+HWq1Gp06dHPYlICAAOp2uwvzqolKpnDqcVSmtG9DIS/pzaZEUkgotQHEugGKgOEua1G5SQNLpAW0jh5B0K1euXMHx48fxySefoE+fPgCAHTt2KMu7dOmCTz/9FNnZ2ZX2Gmk0GthsjmOp/P39AQCXLl1Ct27dAMBhIDYgXUvq/vvvx4svvqjMKz/gm4iI6E5z+ld7+PDhyMrKwrRp02AymRAeHo5NmzYpg6fPnj3rcCuF+++/H19++SXefPNN/OMf/0Dr1q2xZs0ahzDy2muvIS8vD/Hx8TCbzejduzc2bdrkcH+qZcuWYfz48RgwYADUajWGDh2KefPmObTt0Ucfxe+//648L/tBrjd3unfVAo0CpMlWKvUiFZqBohzAXgLkX5YmlYs0eNtdLz2qb9575OPjg8aNG2Px4sUICgrC2bNnMXnyZGV5bGws3nnnHcTExGDmzJkICgrC/v37ERwcjMjISISGhiI9PR0HDhxA06ZN4eXlBXd3d9x3332YNWsWwsLCkJmZiTfffNNhu61bt8b//vc//PDDDwgLC8MXX3yBvXv31kjvHRERUZUIcorFYhEAhMViqbCsoKBApKWliYKCgjvbKJtNiAKzEFd/F+LSISEu7Ls2XTwohOW8ECVFN32JxMRE0b59e6HVakWXLl3E1q1bBQDx7bffCiGEOHPmjBg6dKjw9vYWHh4eokePHiIlJUUIIURhYaEYOnSoMBgMAoBYsmSJEEKItLQ0ERkZKdzd3UV4eLj48ccfBQCxZcsWZb3Ro0cLvV4vDAaDeOGFF8TkyZNF165dlXaNGjVKDBkypJrfsKqptc+TiIiq3c1+v8tTCVFfulPuDKvVCr1eD4vFAm9vb4dlhYWFSE9PR1hYWO3djV0IoDgPKLIABWbAVnxtmc4g9TZpPGunbZWxlUqXI1DVvdv21YnPk4iIqsXNfr/LazCn6zcYKpU0xkjbCPAKlg635WZKY5IKzdLk5gk08peCkhNjkapNaZEU2gquSgPK1a6Apz/g6Sf9mYiIqJbwV6g+U6muDcguzgfysqQwUpIHXM0DXDRSGPFoXPOBxFZ8LQyV5Dsus5cCOZeA3AypLZ7+0ngqIiKiO4zBqKHQeACa5oB3MJAnD9K2FQPWi0COSQ4kftIlAKqLrUTqoSq4Kh3ec2hPI8DdRwptRTlSr1ZpgRTe8rIAnY982M+j+tpDRER0CwxGDY2LG+AdBDQKBAqypRBSWngtkGj10mE2jXOn/CtspeXCUK7jMjdPKQy5G6R2lPHwleaXBaTiHKDwqjRpGklt1XrVzmE/IiJqUBiMGiq1+tphtKIcKRQVWaVB20UWwNVdCkjuPrceGG0vla6tVGCWXgvlxvO7ecg9QwbAVXPj11CpAJ23NBXnA3mZ18JVdq7Uk9UooGrtISIiuk0MRg1d+UBSIvcc5WdLh7XMZ6VDbZ7+UoAq38tjt8kXmzQDhVY4hCFXd6lXyN3n9sYKaTwATSjgFSS354rUq2U+C1gvSYHNw++W12ciIiJyFoMRXeOmAwwhUiDJvyyNRbKXSAOjc0zSIS9No2u3KBH2a+u6auWeIR/pdaqDqxbQNwW8jFJb8rKk9lgvAjnyQO1G/tIg8ruF3Qac3Q1cTQcCOwGBHR0DJxER1SoGI6rIxVUKI40CpMNjeVnSmWT5V6RJqdPIY4Z8pENdNTUGSF2uPflXpcNspYXSY16WtP1GAYCb8/d0uyPsduD8XuDoN8DRNUCu6doyVx1g7AI06S5NTbsDPmEcT0VEVEsYjOjGVOprA6OL8+SB2kXSQGh3HymI3MkfcJUa8Gwstan89ZkKsqVJ6y2fyXabA8erkxDAhX3XwpD1/LVlOj0Q2BnIOCz1vp3fI01l3H2uBaWyydPvju8CEVFDxGBEt1b+opF/UGhoKCZOnIiJEyfKL63Ct99+q9wwuMrtUa7PlCcFpEKzPHjcKo1x0jaSBn5rPAAXbYWgNH36dKxZs6bCjW3/ECEA0yHgyDfA0W8B87X79kHjBbR7FOj4J6Dlg9JAdCGA7NPAhVRpOv+LtH7BVeDkT9JUxtDMMSgFda1bVzAnIqonGIyoVl26dAk+Pj5Vqq00zGg8Ad8wqScrN/PawPHSgms1KhcpILl5SPVu1XxtpIw0qWfoyDdA9qlr8908gDYDgU5/Alo9VHHslUoFNG4pTV2GSfNKi4GMI3JY2ic9Xj4uDTw3n5UCV9k+BXQAmtxzLSwFdJDONiQiotvGYEROKy4uhkZTPQOejUZjtbwOXLXXBo4XWaUxUcX50qOwSZcRKMq5Vp+bKV3gMjdDCjBuHs6d5Zb1m3yY7Fsg69dy7dABrR8GOj4BtIl2vlfHVSOHnXuuzSu0ABcPOPYs5ZqkQ3EZh4F9n0t1jVsDff4GdP4zB3QTEd0m/veS0L9/f4wfPx7jx4+HXq+Hn58fpk6dirL7C4eGhuKtt97CM888A29vb8THxwMAduzYgT59+sDd3R0hISGYMGEC8vKuXeE6MzMTgwcPhru7O8LCwrBs2bIK21apVFizZo3y/Pz584iNjYWvry88PT3Ro0cPpKSkYOnSpZgxYwYOHjwIlUoFlUqFpUuXAgDMZjOeffZZ+Pv7w9vHFw8+9mccPHMF8G8DBHUB/Npi1qffIDD8YXi16Y24v81AYWGBdIaY9SJw5aR0CCvzmNQrk3dZClXlz7oDpMNe2+YCC3sB8+8Fts6UQpGLBmj7KPCnT4G/nwSGfyH1ElXXoS6dHmjRD+gzCXhyGfDqcWDSMWD4/wG9XwHC+koXz7xyAljzPPBhd+CXJVIvGhEROYU9RjVJiIr3BbtT3DycGoD8+eefIy4uDnv27MEvv/yC+Ph4NGvWDM899xwAYO7cuZg2bRoSEhIAAKdOncLAgQPx9ttv47PPPkNWVpYSrpYsWQIAGD16NC5evIgtW7bAzc0NEyZMQGZm5g3bkJubi379+qFJkyb47rvvYDQasW/fPtjtdgwfPhxHjhzBpk2b8NNP0tgbvV4PAPjLX/4Cd3d3bNy4EXq9Hh9//DEGDBiA3377Db6+vli1Zj2mvzMH8+fPR+/evfHF/z7HvA8/QovQZtfuI2cvkc50Ky0EIJ95V6oCcs3AlpXA2S3ApQPXGqt2BVo8IAWgto9K1226k7yDpan9YOl5UQ6w979A8kfS2Kb1E4GfZwO9XgbueYa3ViEiqiKVKOsWoCqxWq3Q6/WwWCzw9vZ2WFZYWIj09HSEhYVBp9NJA4PfCa6dhv7jYpV7LPr374/MzEwcPXoUKjlMTZ48Gd999x3S0tIQGhqKbt264dtvv1XWefbZZ+Hi4oKPP/5Ymbdjxw7069cPeXl5OHv2LNq2bYs9e/bg3nvvBQD8+uuvaN++Pf79739XOvh68eLFePXVV3HmzBn4+vpWaGdlY4x27NiBQYMGITMzE1rttYtJtmrVCq+99hri4+Nx//33o1u3bpg/f76y/L777kNhYeG117KVSJ9XuUNwhSWlSL+QhbCdf4Mu95x0VlxYX2kAdfvB0tlxdU1xvnRobed/pOtPAdIFOu9/CejxV+mMQiKiBuhmv9/lsceIAEhBQVWuhykyMhLvvfcebDYbAKBHjx4O9QcPHsShQ4ccDo8JIWC325Geno7ffvsNrq6u6N69u7K8Xbt2MBgMN2zDgQMH0K1bt0pD0Y0cPHgQubm5aNy4scP8goICnDolDYQ+duwYnn/+eYflkZGR2LJly7UZLm7y1boNZTsD5FqBq3YgfASg9wfaD5EuKFmXaTyA+16QQtCBZcCOf0uHBxOnSX++70WgZ/yd7+EiIrpLMBjVJDcPqeemtrZdjTw9HXufcnNzMXbsWEyYMKFCbbNmzfDbb785vQ13d+cv0Jibm4ugoCBs3bq1wrKbhbBbUqkAN63U69b7FUBXTVfzvlNctVI46jYSOLwa2P6eNJZqy7+AXR8CPZ+TQhKvj3RNbhZwMlG6aXHzXtV3BXciuqswGNUklequudZMSkqKw/Pdu3ejdevWcHGp/Eyte+65B2lpaWjVqlWly9u1a4fS0lKkpqYqh9KOHz8Os9l8wzZ06dIFn376KbKzsyvtNdJoNEoPVvl2mEwmuLq6IjQ0tNLXbd++PVJSUvDMM8847F+D4OIGhD8FdBkOpK2RBo9npklBafdCKTzd/5J0ZfGaUpQDXD4h3QalcWvA2Ln2L8BZprQYOPEjcOBL4MQP0g2RAelaWKG9gFZRQMsBgF/rutNmIqpRDEYEADh79iwmTZqEsWPHYt++ffjwww/x3nvv3bD+9ddfx3333Yfx48fj2WefhaenJ9LS0pCYmIiPPvoIbdu2xcCBAzF27FgsXLgQrq6umDhx4k17hWJjY/HOO+8gJiYGM2fORFBQEPbv34/g4GBERkYiNDQU6enpOHDgAJo2bQovLy9ERUUhMjISMTExmD17Ntq0aYOLFy/i+++/xxNPPIEePXrg5ZdfxujRo9GjRw/06tULy5Ytw9GjR9GiRYuaeCvrJrUL0Gko0OEJ4LeNwLY5wMX90mDtPZ8A94yUBmobmt3e6wshjWm6/JsUgi7/Jk1ZvwE51/Wa+oRJY7Q6DJGuv1QbgcN0WApDh1Y63ubG2EU6KzHnouNFNvXNgFYPSkEprJ9002UiqpcYjAgA8Mwzz6CgoAA9e/aEi4sLXn75ZeW0/Mp06dIFP//8M9544w306dMHQgi0bNkSw4cPV2qWLFmCZ599Fv369UNgYCDefvttTJ069YavqdFo8OOPP+Jvf/sbHn30UZSWlqJDhw7KoOmhQ4fim2++wQMPPACz2YwlS5Zg9OjR2LBhA9544w2MGTMGWVlZMBqN6Nu3LwIDAwEAw4cPx6lTp/Daa6+hsLAQQ4cOxQsvvIAffvihmt69u4haDbQbJJ1JdyoJ+HkOcG43sPdTIHUp0PVJoPck6aKTlSktli5bcPk36cKTSgg6Id2e5UY8A6TrTGUclXqOds2TJu8mUkhq/zjQ7D7nriXlrLzL0mHFA8ukYFSmUaDUoxb+FBDQXgp5mcek9+fkT8DvuwDLWen9SV0qnZHYtCfQaoA0GbvywppE9QjPSnOSU2el3SX69++P8PBwfPDBB7XdlDrlbv08nSIEcGaH1IOU/rM0T6UGOv0Z6DocsF7XC3T1jHTBzMqoXKSrkPu1uW5qJd3/DQCKcqVxPGnfSYewyocpT3+g3WNSUArrWz0XqbSVACcSpTD02w/SZRkA+dpTjwDhT0u3aHG5yf8Ri/OAMzulkHQqSRqrVZ6Hn/QaraKkx7o+QJ+ogarqWWkMRk5iMGo47tbP87ad2yONQTpxi540rbc05savTbnHNtIhMlcnroheUgic3iKFpOPfS1f4LqMzSL1aHR6Xrhfl7EDojKPA/mXyobLL1+YHd5POMuw09PYvt5CdLvcmbZbC5PU9ZUFdr41NCunJq5AT1RE8XZ+InBPSExixCrh0ENj+vvSo9ACVC0CNAqtnXJCbTuq1afuI1LOTvg049h3w6/dAXhZw8Etp0jSSbq/S/nGg9UM3PqEh7wpw5Cupd+jSwWvzPf2vHSoL7PjH2+0bBvg+C9z7rHRo8fyea+ORTIelbV86KA1w13hJVy1v95i0D3Xx2ldE5IA9Rk6qjz1GVDl+nrXEbgPOJks9ScfWOQ7ednWXxvW0fxxoO1C6FcrJn6QwdHzjtUNlajdpefjTUv2d6rXJyQBObZZ7lJKAguxry1Qu0plu7QZL47z0Te5Mm4gIAA+l1RgGo4aDn2cdYLcDF/cBaWul3qSrZ64tU7tJZ4ddf1ZZt6elMVKejSu83B1lt0m3kTm+SeoFyzzquDz4HqD9Y1Jvkn/bWmkiUUPCYFRDGIwaDn6edYwQ0qGqY99JvUmXj0vzPfzkQ2Wx0jWS6qorp6SA9Ot6aTwXyv3T27i1HJIGS+OgeJYbUbVjMKohVQlGoaGht3UVZ6pbCgoKcObMGQajuirrNyAvEwiJuPsGOOdkAMc3SCHp9M/XDgECgFcw0O5RqScptPfdt29EdRSDUQ252RtbUlKCkydPIjg4WLnzO929rly5gszMTLRp0+aGVwAn+sMKLdIlBX5dLz2WP8tNpwfaDJRCUqsBd82V9InqIgajGnKzN1YIgbNnz6KkpATBwcFQszv8riSEQH5+PjIzM2EwGBAUFFTbTaKGoqRQugTAr+uBXzc4XmrA1V26TlLZuCRefZvIKQxGNeRWb2xxcTHS09Nht9troXVUnQwGA4xGI1S8RxbVBrsNOJcCHFsP/LoOMJ+9tkyrB+6Nk24EzAtKElUJg1ENqcoba7fbUVxcfIdbRtXJzc2Nh8+o7hACyDgihaQjXwNXTkjzXXVAt5HSjYB9mtduG4nqOAajGlLVN5aIqEbY7dLA7R3vAxdSpXkqF6DzX4DeE6X7vRFRBQxGNYTBiIjqBCGkq4XveB84vfXa/LaPSjcCDrm31ppGVBcxGNUQBiMiqnMupAI7/i0daiu7PlJoH6D3K9KAbY6TI2IwqikMRkRUZ2X9Buz8D3BoBWAvleYFhUsBqf1gQM1xc9RwMRjVEAYjIqrzLOeBXR8B+z4HSvKleY1bAb0mSlcJd9XUavOIagODUQ1hMCKiu0beFSBlEbDnY+lCkoB0Ze37xwP3jAK0jWq3fUR3EINRDWEwIqK7TlEO8MsSIHk+kGuS5rn7ABHPAz3jAQ/f2m0f0R1Q1d/v27o08/z58xEaGgqdToeIiAjs2bPnpvWrV69Gu3btoNPp0LlzZ2zYsMFhuRAC06ZNQ1BQENzd3REVFYUTJ0441GRnZ2PEiBHw9vaGwWBAXFwccnOvXTq/sLAQo0ePRufOneHq6oqYmJgK7fjmm2/w0EMPwd/fH97e3oiMjMQPP/xwO28BEdHdQ+sF9JoATDwEDP4P4NsCKLgKbJ0J/LsTsOkfQHZ6bbeSqE5wOhitXLkSkyZNQkJCAvbt24euXbsiOjoamZmZldbv2rULsbGxiIuLw/79+xETE4OYmBgcOXJEqZk9ezbmzZuHRYsWISUlBZ6enoiOjkZhYaFSM2LECBw9ehSJiYlYv349tm3bhvj4eGW5zWaDu7s7JkyYgKioqErbsm3bNjz00EPYsGEDUlNT8cADD2Dw4MHYv3+/s28DEdHdx1ULdB8NjP8F+PNnQGBnoCQP2D0fmNcN+HI4cPIn6VpJRA2U04fSIiIicO+99+Kjjz4CIF3lOSQkBC+99BImT55coX748OHIy8vD+vXrlXn33XcfwsPDsWjRIgghEBwcjL/97W949dVXAQAWiwWBgYFYunQpnnzySRw7dgwdOnTA3r170aNHDwDApk2b8Oijj+L8+fMIDg522Obo0aNhNpuxZs2aW+5Px44dMXz4cEybNq1K+89DaURUbwghBaHdC4FTSdfmN24F3PscEB4r3ciWqB6okUNpxcXFSE1NdeiRUavViIqKQnJycqXrJCcnV+jBiY6OVurT09NhMpkcavR6PSIiIpSa5ORkGAwGJRQBQFRUFNRqNVJSUpzZBQd2ux05OTnw9b3x8fWioiJYrVaHiYioXlCpgNYPASO/AcanAhEvAFpv4MpJYNPrwHvtgfWTgMxjtd1SojvGqWB0+fJl2Gw2BAYGOswPDAyEyWSqdB2TyXTT+rLHW9UEBAQ4LHd1dYWvr+8Nt1sVc+fORW5uLoYNG3bDmpkzZ0Kv1ytTSEjIbW+PiKjO8msFPDILmJQGDHoP8G8nHWb75b/AgvuAzwcDx9YBttLabilRjbqtwdf1wZdffokZM2Zg1apVFUJXeVOmTIHFYlGmc+fO3cFWEhHdYVov4N5ngRd3A6PWSReGVKml24+sfBr4T1dg+3tA3uXabilRjXAqGPn5+cHFxQUZGRkO8zMyMmA0Gitdx2g03rS+7PFWNdcP7i4tLUV2dvYNt3szK1aswLPPPotVq1bdcKB2Ga1WC29vb4eJiKjeU6mAsL7A8P8DXj4E9Pkb4NEYsJ4Hkv4JvN8e+Pb5azeyJaonnApGGo0G3bt3R1LStUF6drsdSUlJiIyMrHSdyMhIh3oASExMVOrDwsJgNBodaqxWK1JSUpSayMhImM1mpKZe+wJu3rwZdrsdERERzuwCli9fjjFjxmD58uUYNGiQU+sSETVIhhBgwDTglTQgZhEQ3A2wFQMHlwOfPAh8MgA4uBIoLartlhL9ccJJK1asEFqtVixdulSkpaWJ+Ph4YTAYhMlkEkIIMXLkSDF58mSlfufOncLV1VXMnTtXHDt2TCQkJAg3Nzdx+PBhpWbWrFnCYDCItWvXikOHDokhQ4aIsLAwUVBQoNQMHDhQdOvWTaSkpIgdO3aI1q1bi9jYWIe2HT16VOzfv18MHjxY9O/fX+zfv1/s379fWb5s2TLh6uoq5s+fLy5duqRMZrO5yvtvsVgEAGGxWJx964iI6o9zvwjxdbwQ//QTIsFbmt5tIcRP/xTCfL62W0dUQVV/v2/rytcfffQR5syZA5PJhPDwcMybN0/puenfvz9CQ0OxdOlSpX716tV48803cebMGbRu3RqzZ8/Go48+Wj6cISEhAYsXL4bZbEbv3r2xYMECtGnTRqnJzs7G+PHjsW7dOqjVagwdOhTz5s1Do0bXLmkfGhqK33//vbLwp7Tt559/rrB81KhRDu29GZ6uT0RUTm4WsG8psPczIOeiNE/lIp3yr28CeAcD3k2lR30TwFuex8sA0B3GW4LUEAYjIqJK2EqB498Dez4Bzmy/db3Gq1xYqiw8NQF0/DeWqg+DUQ1hMCIiuoWrvwPZpwHrRcB6QZ4uAhb5z4Xmqr2OxutacHL3BYQdsJcCdpv8WHr7z4VNGkyuDwH0TctNZc+bSGfoUb1R1d9v1zvYJiIiagh8mkvTjRTlAjmXAMv5m4en4hwg61dpqgnFuYC54vALhc5wXXBq4vi8kRFw4c9ofcNPlIiI7ixtI0DbGvBrfeOa68NToRlQuwFqF3lyLTc5+1z+6cu7DFjOSdtwmM5J2yubMg5X3kaVi3z4Tw5KAe2B7mMAjxvfTYHqPh5KcxIPpRERNQBFOVLvVVlQuj44WS9Ih+Wup/UG7nsRiHyRA8zrGI4xqiEMRkREBLsNyM0sF5zOAYdWX+td0hmAXhOAnmOlHjKqdQxGNYTBiIiIKmW3A8e+A7a8A1w+Ls3z8AN6vwLcGwe4uddu+xq4qv5+N9h7pREREVUrtRroGAO8mAz86RPAtwWQfxn48Q3gP+FAymJeHfwuwGBERERUndQuQJdhwLi9wOMfAfpmQK4J2Ph3YN49QOpSwFZS262kG2AwIiIiqgkursA9I4GXUoFB7wFeQdJNeNe9DHzUAziwXBqrRHUKgxEREVFNctUA9z4LTNgPRM8EPP2Bq2eANc8D8yOAI19L45OoTmAwIiIiuhPc3KXT+F8+CETNANx9gCsngK/+CizqDRxbD/B8qFrHYERERHQnaTyB3hOBlw8B/f8hXfso8yiwcgSwuD/w248MSLWIwYiIiKg26LyB/q8DEw8BfV4F3DyBSweAL/8C/Pdh4PRWBqRawOsYOYnXMSIiohqRdxnY+QGw5xOgtFCa590UaNodaHqvNAV15fWQbhMv8FhDGIyIiKhG5ZiA7e8DqUsAW7HjMrUrENhJDko9pEffFoBKVTttvYswGNUQBiMiIrojinKBi/uB83uBC6nSY25GxTp3H6CJHJKadgeadJfmkQMGoxrCYERERLVCCOmebOd/kaYLvwAXDwC2Sq6m7ddGDkvyFNBRuq5SA8ZgVEMYjIiIqM4oLZZuXHte7lE6vxe4ml6xzs0DCO4mjVEyNAP0IYAhRHp092kQh+IYjGoIgxEREdVpeZevHXo7/4v05yLrjevdPOWQ1NQxMJX9uZGxXvQ2MRjVEAYjIiK6q9jt0oUkz+8FMo9Jh+PM5wDLeSAv89brq1wA7yaVhCf5zy4aAEI61CeE/Ge7/NxebpndcVn5da6vM3aWLmdQjar6+333R0AiIiK6MbUa8G8rTdcrKQAsF6SwpAQmOTSZzwLWC4C9FLCclaY75dkkaWxULWAwIiIiaqjc3AG/VtJUGbtNOhNOCUzXhSeLHJxUammckkoFQH5UqSv5s7pcjRpQofI6F80degMqYjAiIiKiyqldAO9gaUJEbbfmjuAtQYiIiIhkDEZEREREMgYjIiIiIhmDEREREZGMwYiIiIhIxmBEREREJGMwIiIiIpIxGBERERHJGIyIiIiIZAxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyW4rGM2fPx+hoaHQ6XSIiIjAnj17blq/evVqtGvXDjqdDp07d8aGDRsclgshMG3aNAQFBcHd3R1RUVE4ceKEQ012djZGjBgBb29vGAwGxMXFITc3V1leWFiI0aNHo3PnznB1dUVMTEylbdm6dSvuueceaLVatGrVCkuXLr2dt4CIiIjqIaeD0cqVKzFp0iQkJCRg37596Nq1K6Kjo5GZmVlp/a5duxAbG4u4uDjs378fMTExiImJwZEjR5Sa2bNnY968eVi0aBFSUlLg6emJ6OhoFBYWKjUjRozA0aNHkZiYiPXr12Pbtm2Ij49XlttsNri7u2PChAmIioqqtC3p6ekYNGgQHnjgARw4cAATJ07Es88+ix9++MHZt4GIiIjqI+Gknj17inHjxinPbTabCA4OFjNnzqy0ftiwYWLQoEEO8yIiIsTYsWOFEELY7XZhNBrFnDlzlOVms1lotVqxfPlyIYQQaWlpAoDYu3evUrNx40ahUqnEhQsXKmxz1KhRYsiQIRXmv/baa6Jjx44O84YPHy6io6NvsdfXWCwWAUBYLJYqr0NERES1q6q/3071GBUXFyM1NdWhR0atViMqKgrJycmVrpOcnFyhByc6OlqpT09Ph8lkcqjR6/WIiIhQapKTk2EwGNCjRw+lJioqCmq1GikpKVVu/63aUpmioiJYrVaHiYiIiOonp4LR5cuXYbPZEBgY6DA/MDAQJpOp0nVMJtNN68seb1UTEBDgsNzV1RW+vr433K4zbbFarSgoKKh0nZkzZ0Kv1ytTSEhIlbdHREREdxeelXYLU6ZMgcViUaZz587VdpOIiIiohjgVjPz8/ODi4oKMjAyH+RkZGTAajZWuYzQab1pf9nirmusHd5eWliI7O/uG23WmLd7e3nB3d690Ha1WC29vb4eJiIiI6iengpFGo0H37t2RlJSkzLPb7UhKSkJkZGSl60RGRjrUA0BiYqJSHxYWBqPR6FBjtVqRkpKi1ERGRsJsNiM1NVWp2bx5M+x2OyIiIqrc/lu1hYiIiBo4Z0d1r1ixQmi1WrF06VKRlpYm4uPjhcFgECaTSQghxMiRI8XkyZOV+p07dwpXV1cxd+5ccezYMZGQkCDc3NzE4cOHlZpZs2YJg8Eg1q5dKw4dOiSGDBkiwsLCREFBgVIzcOBA0a1bN5GSkiJ27NghWrduLWJjYx3advToUbF//34xePBg0b9/f7F//36xf/9+Zfnp06eFh4eH+Pvf/y6OHTsm5s+fL1xcXMSmTZuqvP88K42IiOjuU9Xfb6eDkRBCfPjhh6JZs2ZCo9GInj17it27dyvL+vXrJ0aNGuVQv2rVKtGmTRuh0WhEx44dxffff++w3G63i6lTp4rAwECh1WrFgAEDxPHjxx1qrly5ImJjY0WjRo2Et7e3GDNmjMjJyXGoad68uQBQYSpvy5YtIjw8XGg0GtGiRQuxZMkSp/adwYiIiOjuU9Xfb5UQQtRad9VdyGq1Qq/Xw2KxcLwRERHRXaKqv988K42IiIhIxmBEREREJGMwIiIiIpIxGBERERHJGIyIiIiIZAxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQkYzAiIiIikjEYEREREckYjIiIiIhkDEZEREREMgYjIiIiIhmDEREREZGMwYiIiIhIxmBEREREJGMwIiIiIpIxGBERERHJGIyIiIiIZAxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIi2W0Fo/nz5yM0NBQ6nQ4RERHYs2fPTetXr16Ndu3aQafToXPnztiwYYPDciEEpk2bhqCgILi7uyMqKgonTpxwqMnOzsaIESPg7e0Ng8GAuLg45ObmOtQcOnQIffr0gU6nQ0hICGbPnl2hLR988AHatm0Ld3d3hISE4JVXXkFhYeHtvA1ERERU3wgnrVixQmg0GvHZZ5+Jo0ePiueee04YDAaRkZFRaf3OnTuFi4uLmD17tkhLSxNvvvmmcHNzE4cPH1ZqZs2aJfR6vVizZo04ePCgePzxx0VYWJgoKChQagYOHCi6du0qdu/eLbZv3y5atWolYmNjleUWi0UEBgaKESNGiCNHjojly5cLd3d38fHHHys1y5YtE1qtVixbtkykp6eLH374QQQFBYlXXnmlyvtvsVgEAGGxWJx524iIiKgWVfX32+lg1LNnTzFu3Djluc1mE8HBwWLmzJmV1g8bNkwMGjTIYV5ERIQYO3asEEIIu90ujEajmDNnjrLcbDYLrVYrli9fLoQQIi0tTQAQe/fuVWo2btwoVCqVuHDhghBCiAULFggfHx9RVFSk1Lz++uuibdu2yvNx48aJBx980KEtkyZNEr169ary/jMYERER3X2q+vvt1KG04uJipKamIioqSpmnVqsRFRWF5OTkStdJTk52qAeA6OhopT49PR0mk8mhRq/XIyIiQqlJTk6GwWBAjx49lJqoqCio1WqkpKQoNX379oVGo3HYzvHjx3H16lUAwP3334/U1FTl0N/p06exYcMGPProozfc56KiIlitVoeJiIiI6idXZ4ovX74Mm82GwMBAh/mBgYH49ddfK13HZDJVWm8ymZTlZfNuVhMQEODYcFdX+Pr6OtSEhYVVeI2yZT4+Pnjqqadw+fJl9O7dG0IIlJaW4vnnn8c//vGPG+7zzJkzMWPGjBsuJyIiovqjQZ2VtnXrVrzzzjtYsGAB9u3bh2+++Qbff/893nrrrRuuM2XKFFgsFmU6d+7cHWwxERER3UlO9Rj5+fnBxcUFGRkZDvMzMjJgNBorXcdoNN60vuwxIyMDQUFBDjXh4eFKTWZmpsNrlJaWIjs72+F1KttO+W1MnToVI0eOxLPPPgsA6Ny5M/Ly8hAfH4833ngDanXFnKjVaqHVam/wjhAREVF94lSPkUajQffu3ZGUlKTMs9vtSEpKQmRkZKXrREZGOtQDQGJiolIfFhYGo9HoUGO1WpGSkqLUREZGwmw2IzU1VanZvHkz7HY7IiIilJpt27ahpKTEYTtt27aFj48PACA/P79C+HFxcQEgXTKAiIiIGjhnR3WvWLFCaLVasXTpUpGWlibi4+OFwWAQJpNJCCHEyJEjxeTJk5X6nTt3CldXVzF37lxx7NgxkZCQUOnp+gaDQaxdu1YcOnRIDBkypNLT9bt16yZSUlLEjh07ROvWrR1O1zebzSIwMFCMHDlSHDlyRKxYsUJ4eHg4nK6fkJAgvLy8xPLly8Xp06fFjz/+KFq2bCmGDRtW5f3nWWlERER3nxo7XV8IIT788EPRrFkzodFoRM+ePcXu3buVZf369ROjRo1yqF+1apVo06aN0Gg0omPHjuL77793WG6328XUqVNFYGCg0Gq1YsCAAeL48eMONVeuXBGxsbGiUaNGwtvbW4wZM0bk5OQ41Bw8eFD07t1baLVa0aRJEzFr1iyH5SUlJWL69OmiZcuWQqfTiZCQEPHiiy+Kq1evVnnfGYyIiIjuPlX9/VYJwWNIzrBardDr9bBYLPD29q7t5hAREVEVVPX3u0GdlUZERER0MwxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQkYzAiIiIikjEYEREREckYjIiIiIhkDEZEREREMgYjIiIiIhmDEREREZGMwYiIiIhIxmBEREREJGMwIiIiIpIxGBERERHJGIyIiIiIZAxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQku61gNH/+fISGhkKn0yEiIgJ79uy5af3q1avRrl076HQ6dO7cGRs2bHBYLoTAtGnTEBQUBHd3d0RFReHEiRMONdnZ2RgxYgS8vb1hMBgQFxeH3Nxch5pDhw6hT58+0Ol0CAkJwezZsyu0xWw2Y9y4cQgKCoJWq0WbNm0qtIeIiIgaJqeD0cqVKzFp0iQkJCRg37596Nq1K6Kjo5GZmVlp/a5duxAbG4u4uDjs378fMTExiImJwZEjR5Sa2bNnY968eVi0aBFSUlLg6emJ6OhoFBYWKjUjRozA0aNHkZiYiPXr12Pbtm2Ij49XllutVjz88MNo3rw5UlNTMWfOHEyfPh2LFy9WaoqLi/HQQw/hzJkz+Oqrr3D8+HF88sknaNKkibNvAxEREdVHwkk9e/YU48aNU57bbDYRHBwsZs6cWWn9sGHDxKBBgxzmRUREiLFjxwohhLDb7cJoNIo5c+Yoy81ms9BqtWL58uVCCCHS0tIEALF3716lZuPGjUKlUokLFy4IIYRYsGCB8PHxEUVFRUrN66+/Ltq2bas8X7hwoWjRooUoLi52drcVFotFABAWi+W2X4OIiIjurKr+fjvVY1RcXIzU1FRERUUp89RqNaKiopCcnFzpOsnJyQ71ABAdHa3Up6enw2QyOdTo9XpEREQoNcnJyTAYDOjRo4dSExUVBbVajZSUFKWmb9++0Gg0Dts5fvw4rl69CgD47rvvEBkZiXHjxiEwMBCdOnXCO++8A5vNdsN9LioqgtVqdZiIiIiofnIqGF2+fBk2mw2BgYEO8wMDA2EymSpdx2Qy3bS+7PFWNQEBAQ7LXV1d4evr61BT2WuU38bp06fx1VdfwWazYcOGDZg6dSree+89vP322zfc55kzZ0Kv1ytTSEjIDWuJiIjo7tagzkqz2+0ICAjA4sWL0b17dwwfPhxvvPEGFi1adMN1pkyZAovFokznzp27gy0mIiKiO8nVmWI/Pz+4uLggIyPDYX5GRgaMRmOl6xiNxpvWlz1mZGQgKCjIoSY8PFypuX5wd2lpKbKzsx1ep7LtlN9GUFAQ3Nzc4OLiotS0b98eJpMJxcXFDofhymi1Wmi12kr3jYiIiOoXp3qMNBoNunfvjqSkJGWe3W5HUlISIiMjK10nMjLSoR4AEhMTlfqwsDAYjUaHGqvVipSUFKUmMjISZrMZqampSs3mzZtht9sRERGh1Gzbtg0lJSUO22nbti18fHwAAL169cLJkydht9uVmt9++w1BQUGVhiIiIiJqYJwd1b1ixQqh1WrF0qVLRVpamoiPjxcGg0GYTCYhhBAjR44UkydPVup37twpXF1dxdy5c8WxY8dEQkKCcHNzE4cPH1ZqZs2aJQwGg1i7dq04dOiQGDJkiAgLCxMFBQVKzcCBA0W3bt1ESkqK2LFjh2jdurWIjY1VlpvNZhEYGChGjhwpjhw5IlasWCE8PDzExx9/rNScPXtWeHl5ifHjx4vjx4+L9evXi4CAAPH2229Xef95VhoREdHdp6q/304HIyGE+PDDD0WzZs2ERqMRPXv2FLt371aW9evXT4waNcqhftWqVaJNmzZCo9GIjh07iu+//95hud1uF1OnThWBgYFCq9WKAQMGiOPHjzvUXLlyRcTGxopGjRoJb29vMWbMGJGTk+NQc/DgQdG7d2+h1WpFkyZNxKxZsyq0fdeuXSIiIkJotVrRokUL8a9//UuUlpZWed8ZjIiIiO4+Vf39VgkhRO32Wd1drFYr9Ho9LBYLvL29a7s5REREVAVV/f1uUGelEREREd0MgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQkYzAiIiIikjEYEREREckYjIiIiIhkDEZEREREMgYjIiIiIhmDEREREZGMwYiIiIhIxmBEREREJGMwIiIiIpIxGBERERHJGIyIiIiIZAxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQkYzAiIiIikjEYEREREckYjIiIiIhktxWM5s+fj9DQUOh0OkRERGDPnj03rV+9ejXatWsHnU6Hzp07Y8OGDQ7LhRCYNm0agoKC4O7ujqioKJw4ccKhJjs7GyNGjIC3tzcMBgPi4uKQm5vrUHPo0CH06dMHOp0OISEhmD179g3btGLFCqhUKsTExDi380RERFRvOR2MVq5ciUmTJiEhIQH79u1D165dER0djczMzErrd+3ahdjYWMTFxWH//v2IiYlBTEwMjhw5otTMnj0b8+bNw6JFi5CSkgJPT09ER0ejsLBQqRkxYgSOHj2KxMRErF+/Htu2bUN8fLyy3Gq14uGHH0bz5s2RmpqKOXPmYPr06Vi8eHGFNp05cwavvvoq+vTp4+zuExERUX0mnNSzZ08xbtw45bnNZhPBwcFi5syZldYPGzZMDBo0yGFeRESEGDt2rBBCCLvdLoxGo5gzZ46y3Gw2C61WK5YvXy6EECItLU0AEHv37lVqNm7cKFQqlbhw4YIQQogFCxYIHx8fUVRUpNS8/vrrom3btg7bLi0tFffff7/49NNPxahRo8SQIUOc2n+LxSIACIvF4tR6REREVHuq+vvtVI9RcXExUlNTERUVpcxTq9WIiopCcnJypeskJyc71ANAdHS0Up+eng6TyeRQo9frERERodQkJyfDYDCgR48eSk1UVBTUajVSUlKUmr59+0Kj0Ths5/jx47h69aoy75///CcCAgIQFxdXpX0uKiqC1Wp1mIiIiKh+cioYXb58GTabDYGBgQ7zAwMDYTKZKl3HZDLdtL7s8VY1AQEBDstdXV3h6+vrUFPZa5Tfxo4dO/Df//4Xn3zySdV2GMDMmTOh1+uVKSQkpMrrEhER0d2lwZyVlpOTg5EjR+KTTz6Bn59fldebMmUKLBaLMp07d64GW0lERES1ydWZYj8/P7i4uCAjI8NhfkZGBoxGY6XrGI3Gm9aXPWZkZCAoKMihJjw8XKm5fnB3aWkpsrOzHV6nsu2ULTt16hTOnDmDwYMHK8vtdjsAqffp+PHjaNmyZYX2a7VaaLXaSveNiIiI6heneow0Gg26d++OpKQkZZ7dbkdSUhIiIyMrXScyMtKhHgASExOV+rCwMBiNRocaq9WKlJQUpSYyMhJmsxmpqalKzebNm2G32xEREaHUbNu2DSUlJQ7badu2LXx8fNCuXTscPnwYBw4cUKbHH38cDzzwAA4cOMBDZEREROT8WWkrVqwQWq1WLF26VKSlpYn4+HhhMBiEyWQSQggxcuRIMXnyZKV+586dwtXVVcydO1ccO3ZMJCQkCDc3N3H48GGlZtasWcJgMIi1a9eKQ4cOiSFDhoiwsDBRUFCg1AwcOFB069ZNpKSkiB07dojWrVuL2NhYZbnZbBaBgYFi5MiR4siRI2LFihXCw8NDfPzxxzfcF56VRkRE1DBU9ffbqUNpADB8+HBkZWVh2rRpMJlMCA8Px6ZNm5SBzmfPnoVafa0j6v7778eXX36JN998E//4xz/QunVrrFmzBp06dVJqXnvtNeTl5SE+Ph5msxm9e/fGpk2boNPplJply5Zh/PjxGDBgANRqNYYOHYp58+Ypy/V6PX788UeMGzcO3bt3h5+fH6ZNm+ZwrSMiIiKim1EJIURtN+JuYrVaodfrYbFY4O3tXdvNISIioiqo6u93gzkrjYiIiOhWGIyIiIiIZAxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQkYzAiIiIikjEYEREREckYjIiIiIhkDEZEREREMgYjIiIiIhmDEREREZGMwYiIiIhIxmBEREREJGMwIiIiIpIxGBERERHJGIyIiIiIZAxGRERERDIGIyIiIiIZgxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRgRERERyRiMiIiIiGQMRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQku61gNH/+fISGhkKn0yEiIgJ79uy5af3q1avRrl076HQ6dO7cGRs2bHBYLoTAtGnTEBQUBHd3d0RFReHEiRMONdnZ2RgxYgS8vb1hMBgQFxeH3Nxch5pDhw6hT58+0Ol0CAkJwezZsx2Wf/LJJ+jTpw98fHzg4+ODqKioW7adiIiIGg6ng9HKlSsxadIkJCQkYN++fejatSuio6ORmZlZaf2uXbsQGxuLuLg47N+/HzExMYiJicGRI0eUmtmzZ2PevHlYtGgRUlJS4OnpiejoaBQWFio1I0aMwNGjR5GYmIj169dj27ZtiI+PV5ZbrVY8/PDDaN68OVJTUzFnzhxMnz4dixcvVmq2bt2K2NhYbNmyBcnJyQgJCcHDDz+MCxcuOPs2EBERUX0knNSzZ08xbtw45bnNZhPBwcFi5syZldYPGzZMDBo0yGFeRESEGDt2rBBCCLvdLoxGo5gzZ46y3Gw2C61WK5YvXy6EECItLU0AEHv37lVqNm7cKFQqlbhw4YIQQogFCxYIHx8fUVRUpNS8/vrrom3btjfcl9LSUuHl5SU+//zzqu6+sFgsAoCwWCxVXoeIiIhqV1V/v53qMSouLkZqaiqioqKUeWq1GlFRUUhOTq50neTkZId6AIiOjlbq09PTYTKZHGr0ej0iIiKUmuTkZBgMBvTo0UOpiYqKglqtRkpKilLTt29faDQah+0cP34cV69erbRt+fn5KCkpga+v7w33uaioCFar1WEiIiKi+smpYHT58mXYbDYEBgY6zA8MDITJZKp0HZPJdNP6ssdb1QQEBDgsd3V1ha+vr0NNZa9RfhvXe/311xEcHFwhuJU3c+ZM6PV6ZQoJCblhLREREd3dGuxZabNmzcKKFSvw7bffQqfT3bBuypQpsFgsynTu3Lk72EoiIiK6k1ydKfbz84OLiwsyMjIc5mdkZMBoNFa6jtFovGl92WNGRgaCgoIcasLDw5Wa6wd3l5aWIjs72+F1KttO+W2UmTt3LmbNmoWffvoJXbp0uek+a7VaaLXam9YQERFR/eBUj5FGo0H37t2RlJSkzLPb7UhKSkJkZGSl60RGRjrUA0BiYqJSHxYWBqPR6FBjtVqRkpKi1ERGRsJsNiM1NVWp2bx5M+x2OyIiIpSabdu2oaSkxGE7bdu2hY+PjzJv9uzZeOutt7Bp0yaHMUtERERETp+VtmLFCqHVasXSpUtFWlqaiI+PFwaDQZhMJiGEECNHjhSTJ09W6nfu3ClcXV3F3LlzxbFjx0RCQoJwc3MThw8fVmpmzZolDAaDWLt2rTh06JAYMmSICAsLEwUFBUrNwIEDRbdu3URKSorYsWOHaN26tYiNjVWWm81mERgYKEaOHCmOHDkiVqxYITw8PMTHH3/ssB2NRiO++uorcenSJWXKycmp8v7zrDQiIqK7T1V/v50ORkII8eGHH4pmzZoJjUYjevbsKXbv3q0s69evnxg1apRD/apVq0SbNm2ERqMRHTt2FN9//73DcrvdLqZOnSoCAwOFVqsVAwYMEMePH3eouXLlioiNjRWNGjUS3t7eYsyYMRUCzcGDB0Xv3r2FVqsVTZo0EbNmzXJY3rx5cwGgwpSQkFDlfWcwIiIiuvtU9fdbJYQQtdZddReyWq3Q6/WwWCzw9vau7eYQEdFtSLtoRVNfd3jr3Gq7KXSHVPX3u8GelUZERA3Tkp3peHTedjw4dyu+P3QJ7B+g8hiMiIiowUj9PRv/+v4YAOBybjHGfbkPz/9fKjKthbdYkxoKBiMiImoQLucWYdyy/Si1CwzqHIQJD7aCq1qFH45mIOr9n7Hql3PsPSIGIyIiqv9sdoGXV+yHyVqIFv6eePfPXTDp4bZY91JvdG6ih7WwFK99dQjPfLYH57Lza7u5VIsYjIiIqN774KffsPPkFbi7uWDR093RSCtd37h9kDe+ffF+THmkHbSuamw/cRnRH2zDkp3psNnZe9QQMRgREVG9tuV4Jj7cfBIAMPNPndEm0MthuauLGmP7tcTGl/ugZ5gv8ottmLEuDcM+TsbJzJzaaDLVIgYjIiKqt85fzccrKw8AAJ6+rxliujW5YW0L/0ZY8dx9eCumEzw1Lkj9/Soe/c8OzN9yEiU2+x1qMdU2BiMiIqqXikpteHHZPpjzS9C1qR5TH+twy3XUahVG3tccP07qh/5t/VFss2POD8cx5KOdOHLBcgdaTbWNwYiIiOqlt9an4dB5Cwwebpg/4h5oXV2qvG4TgzuWjL4X/x7eFQYPN6RdsmLI/J14d9OvKCyx1WCrqbYxGBERUb2zZv8F/N/us1CpgH8PD0dTHw+nX0OlUuGJbk2R+Eo/DOoSBJtdYOHWU3j0P9ux90x2DbSa6gIGIyIiqld+y8jBlG8OAwBeeqAVHmgb8Idez99Li/lP3YOPR3aHv5cWpy/nYdjHyUhYewS5RaXV0WSqQxiMiIio3sgtKsXz/5eKghIberfyw8tRbarttaM7GvHTK/0wrEdTCAF8nvw7ov+9DT//llVt26Dax2BERET1ghACr399CKez8mD01uE/T4bDRa2q1m3oPdww+89d8X9xEWjq444L5gKM+mwP/rbqIMz5xdW6LaodDEZERFQvLN11Bt8fugRXtQrzR9yDxo20Nbat3q398MPEvhjTKxQqFfD1vvN46N/bsCedY4/udirBG8M4xWq1Qq/Xw2KxwNvbu7abQ39QUakNmdYiZOYUItNahAxrITJyihzmZeUWwd3NBYHeWgR46aRHbx0CvLQI9NYhwFuLQC8dDB5uUKmq93+n5MhSUIJfzmRj9+kr2H06G5k5hWhn9EbnJnp0bqpH5yZ6BOl1/BwaoNTfr2L4x8kotQtMe6wD/to77A5uOxuvfXUIp7Ly4KpWYdrgDhh5X/M69ffQnF+MN9YcwbGLVozpFYph94Y4dZZefVDV328GIycxGN0dKgs8mTlFyCg/L6cQ5vySatumxkUNfy8tA1Q1shaWYG/6tSB09KIFt7pLQ2NPDTo10aNLUz06NWFYagiu5BbhsQ934JKlEIO6BOGj2G53/PPOLy7F618fxrqDFwEAf+neFG/FdILOrfbDx/6zVzH+y/24YC5Q5gXpdXjxgVYY1qNpgwlIDEY1hMGobsovLkViWgbWHbyI1N+v4qoTgUfjqnYMM17XQkyAtxb+XlrkF9uQqYSrsmBVpMzLzqv62AKNixot/D0x+ZF26P8Hz5apb6oShML8PHFfC1/c16IxmhjcceySFYcvWHD4ghW/ZeRUen+rxp4apUeJYal+sdkFRi/Zg+0nLqOFvye+G99buQ/anSaEwCfbT2PWxl9hF0DXpnosGtkdQXr3WmvPf3ekY9bGX1FqF2je2APDeoTgi+TfYbIWAmhYAYnBqIYwGNUdxaV2bD+Rhe8OXsSPRzNQcN1F124VeAK9dQj00sHb3fUP/0AWldqQlVOEzHJhKcNaKPdQSfMyrIUVAtuf7mmCqYM6wMdT84e2f7dyNghFhDWGUa+74esVlthw7JIVRy5YcPiCBYfOW3AiM7fSsOTXSKOEpMrCkhACRaV25BaVIq+oFLlFpcgvtinPpXk25BeVIre4bJ5NrpOW5RWVoqi0ei8GqIIKXZrqMSS8Cfq28av3P2a38n7ib5iXdALubi5YM64X2hq9br1SDdt+IgsvLd8Pc34J/BppsGBEd/QM872jbbDkl+Bvqw/ip2MZAIBBnYMwa2hneOncUFhiw6pfzmHBllMNKiAxGNUQBqPaZbcL7DmTje8OXsSGw5ccDoU18/XAkPBgPNzBiGa+HtUSeKpb2SG+JTvPYMmudAgh/UD/c0gnPNLJWOfaW92shWVjhKQwdOTCHwtCVeFMWGrsqYHGVa2EoLp+d3VvnSse7RyEx8ODERHWuNrPwKrrth7PxJileyEE8O/hXfFEt6a13STF2Sv5iP/iF/xqyoGrWoWEwR3w9B0ad1T+0JnGRY2pgzvg6YhmFbbd0AISg1ENqe/BKLeoVO7dqGx8TiFKbHa0C5IHuzbRo02gFzSuNXtyoxACRy9a8d3Bi1h38CIuWQqVZX6NtBjcNQhDwpuga1P9XRUsUn+/ite/PoSTmbkAgOiOgXhrSCcEeP+xIPBHmSyFWPTzKaSkZ6M6/3kosdmRfjmvxoNQVTgTlgDAQ+MCT60rGmld4al1gafGFZ5aV3le+efl6uR5Wjc1qvNvZUGJDZuPZWLdoYvIsBYp8wO8tBjcNRiPdw1Gl7vsu3A7LpgLMGjedpjzSzAiohn+9UTn2m5SBfnFpXjtq0NYf+gSgJofd1TZobP5T92DTk30N12vLCDN33JS+TtVHwMSg1ENuVuD0a0CT5b8mFfsXLe/xkWNdkFeSlDqVI1hKf1yHr47cBFrD17A6aw8Zb6XzhWPdDJiSHgT3Nfi7v5fclGpDR9tPomFW0+h1C7grXPFm491wF+6N73jP2wmSyEWbj2J5XvOobgG7yReG0GoKgpLbDiRkQsBUS4EucLDzQXqOvh3zGYXSEm/gnUHL+L7Q5dgLbx2BebQxh54PLwJHu8ajFYBjWqxlTWjqNSGYR/vxsFzZnRpqsfq5yPr7I+3EAKLt53Gu5vkcUchBix6+p5qH3dU2aGzmUM7w1vnVuXXKCyxYeXec1iw9VpACpYD0l/qQUBiMKohdzoY2ewCecXXxjTkyeMWcotKkVduHEN+0bU/55arv5Jb7HTgaaR1RYCXFgHlxucEeuvg76WFSqXC0YsW6X/a5y0O/xiX0bio0T7Iy2H8RlujF9xcbh2WMqyFWHfwIr47eBGHzl+7k7XWVY2o9oF4PDwY/dv63/Vf0OulXbTi9a8P4bB89+4+rf3wzhOdEeLr/P2dnFVZIOoZ6osxvULh5cQ/qreiUgEt/RvVmSBUnxSV2rDtt8v47uBFJKaZUFhyLdh2DPbGkPBgPNYlGMGG6h8EXGKz41x2Pk5n5SH9ch5OX87D+av5aOrjjvtaNMZ9LRojsJp7QaetPYL/Jf8Ovbsb1r/U+458T/6o7SeyMP7L/bAUSOOOFj7dHfeGVs+4owqHzh5r/4cO29XXgMRgVENqKhglrD2CY5dylMBTFoCuH1D8RzTSusphRz593Ota4An01inzPKt4RocQAmez8+UzgqSgdOTCDcKSqxrtjZWHJUt+CTYeuYS1By5id/oVlP2NdFGr0LuVH4aEB+OhDoHV+iNdF5Xa7Ph0Rzr+nfgbikrt8NC44LXotngmMrRGeixuFIgmPtQakS0a1/tDMfVVXpF0huZ3By9i229ZKC13eLBnmC8e7xqMRzsHwdeJAf9CCGTmFF0LP1m5Sgg6m51/y7FYLfw8EdGisdJT+EeC0toDF/DyigMAgCWj78UD7e6eMzure9yREAKf7TyDWRuPocRW9UNnVVXfAhKDUQ2pqWA0dOEupP5+9YbLXdUqaQyDPNbBYbxDufEMjbSOy309NU4Hnj/C2bDUws8Tp7JyUWK79tewR3MfDAmX/vGuySvX1lWns3Ix+evD2CPfvbt7cx+8O7RLtR0SYSBqOLLzirHh8CV8d/CiwxWZXdUq9G3jj8e7Sv/pKPu3IaewBGcu5+P05dxyPUC5SM/Ku2mvs7ubC8L8PBHm74mWfp5o6uOBE5k5NzzT8HaD0omMHAyZvxP5xTaMf6AVXo1u6/ybUsuuH3c0rEdT/HOI8+OOLPklePWrg0hMu/1DZ1VVXwISg1ENqalg9PNvWcgtLHUIOuWDj9ZVfdf+YDmEpfNSYLo+LLUzemFIeBMM7hqEpj51v1u8ptntAstSfsesjb8ir9gGjYsaL0e1RnzfFlU6JFkZBqKG7aK5QDlMffSiVZmvc1OjQ5A3zl8tQGZO0Q3XV6uAEF8PhPl5ooVfIyUEhfl7wuh942tCXX+18tsNSnlFpRgyfydOZuaiV6vG+N9fI+7a8YWVjTv6+OnuVT7MfOCcGeOW7au2Q2dVdaOA1CG4enqoypvyaDu09K/e8XEMRjXkbh18XdeUhaXjphyE+nmiTWDtX3ukLrpgLsA/vjms3L27Q5A3Zv+5i1Nd5TcMRFGtEdmSgaghOpmZg+8OSCHpzJV8h2V+jTQO4aeFnyda+Huima9ntZxUcTtBKcBLiwkrDmDdwYsweuuwfkJv+NWD3mTHcUdaLHz6npuOO7r+0FkzXw8sGFF9h86qqrKAVN2+efF+3NPMp1pfk8GohjAY0Z0mhMC3+y/gn+vTYM4vgYtahbF9W2DCgNY37X5nIKJbEULg0HkLzlzJQ/PGngjz84Te/c6O5bs+KB25aMH1v0rBeh0uWgrhqlZh5dj70L35nb1YYk2q6rgjS34J/v7VQfwoHzp7tLMRs4Z2qZFDZ1VVWGLDT8cykFPJUIk/Kqp9IPy9qjf8MhjVEAYjqi1ZOUWY/t1RfH9YGpvQwt8Ts4d2QY/r/ofJQER3M0tBuauhp1/B0YtWJShNfawD4u7gzWHvlPziUvz9q0P4Xh53NLxHCP4Z01EZu3P9obM3H2tf525SezdgMKohDEZU2zYdMWHq2iPIyimCSgU8c19zvDawHXIKSysEontDffBKVBsGIrprWQpKsCc9Gza7HdEd6+/V4YUQ+HjbacyWxx2Fhxiw8Ol7sOGwyeHQ2fyn7kHnpnf20Fl9wWBUQxiMqC6w5Jfg7e/TsDr1PADpqsfm/BIGIqK73LbfpPusWQpKoHFVo7hU+k7XhUNndzsGoxrCYER1yfYTWZjyzWGcv1oAgIGIqD4oP+6Ih86qD4NRDWEworomr6gUq345h7ZGL552T1RP5BeXYvUv59EzzBftg/hbUx0YjGoIgxEREdHdp6q/3zV7W3QiIiKiuwiDEREREZGMwYiIiIhIxmBEREREJGMwIiIiIpIxGBERERHJbisYzZ8/H6GhodDpdIiIiMCePXtuWr969Wq0a9cOOp0OnTt3xoYNGxyWCyEwbdo0BAUFwd3dHVFRUThx4oRDTXZ2NkaMGAFvb28YDAbExcUhNzfXoebQoUPo06cPdDodQkJCMHv2bKfbQkRERA2X08Fo5cqVmDRpEhISErBv3z507doV0dHRyMzMrLR+165diI2NRVxcHPbv34+YmBjExMTgyJEjSs3s2bMxb948LFq0CCkpKfD09ER0dDQKCwuVmhEjRuDo0aNITEzE+vXrsW3bNsTHxyvLrVYrHn74YTRv3hypqamYM2cOpk+fjsWLFzvVFiIiImrAhJN69uwpxo0bpzy32WwiODhYzJw5s9L6YcOGiUGDBjnMi4iIEGPHjhVCCGG324XRaBRz5sxRlpvNZqHVasXy5cuFEEKkpaUJAGLv3r1KzcaNG4VKpRIXLlwQQgixYMEC4ePjI4qKipSa119/XbRt27bKbakKi8UiAAiLxVLldYiIiKh2VfX326keo+LiYqSmpiIqKkqZp1arERUVheTk5ErXSU5OdqgHgOjoaKU+PT0dJpPJoUav1yMiIkKpSU5OhsFgQI8ePZSaqKgoqNVqpKSkKDV9+/aFRqNx2M7x48dx9erVKrWlMkVFRbBarQ4TERER1U9OBaPLly/DZrMhMDDQYX5gYCBMJlOl65hMppvWlz3eqiYgIMBhuaurK3x9fR1qKnuN8tu4VVsqM3PmTOj1emUKCQm5YS0RERHd3XhW2i1MmTIFFotFmc6dO1fbTSIiIqIa4lQw8vPzg4uLCzIyMhzmZ2RkwGg0VrqO0Wi8aX3Z461qrh/cXVpaiuzsbIeayl6j/DZu1ZbKaLVaeHt7O0xERERUP7k6U6zRaNC9e3ckJSUhJiYGAGC325GUlITx48dXuk5kZCSSkpIwceJEZV5iYiIiIyMBAGFhYTAajUhKSkJ4eDgA6QyzlJQUvPDCC8prmM1mpKamonv37gCAzZs3w263IyIiQql54403UFJSAjc3N2U7bdu2hY+PT5XaUhVCCKWNREREdHco+90u+x2/IWdHda9YsUJotVqxdOlSkZaWJuLj44XBYBAmk0kIIcTIkSPF5MmTlfqdO3cKV1dXMXfuXHHs2DGRkJAg3NzcxOHDh5WaWbNmCYPBINauXSsOHTokhgwZIsLCwkRBQYFSM3DgQNGtWzeRkpIiduzYIVq3bi1iY2OV5WazWQQGBoqRI0eKI0eOiBUrVggPDw/x8ccfO9WWWzl37pwAwIkTJ06cOHG6C6dz587d9HfeqR4jABg+fDiysrIwbdo0mEwmhIeHY9OmTcqg5rNnz0KtvnaE7v7778eXX36JN998E//4xz/QunVrrFmzBp06dVJqXnvtNeTl5SE+Ph5msxm9e/fGpk2boNPplJply5Zh/PjxGDBgANRqNYYOHYp58+Ypy/V6PX788UeMGzcO3bt3h5+fH6ZNm+ZwraOqtOVWgoODce7cOXh5eUGlUjn79t2Q1WpFSEgIzp07x8N1dQw/m7qJn0vdxc+mbmron4sQAjk5OQgODr5pnUqIW/Up0Z1gtVqh1+thsVga5F/YuoyfTd3Ez6Xu4mdTN/FzqRqelUZEREQkYzAiIiIikjEY1RFarRYJCQnQarW13RS6Dj+buomfS93Fz6Zu4udSNRxjRERERCRjjxERERGRjMGIiIiISMZgRERERCRjMCIiIiKSMRjVEfPnz0doaCh0Oh0iIiKwZ8+e2m5SgzZ9+nSoVCqHqV27drXdrAZp27ZtGDx4MIKDg6FSqbBmzRqH5UIITJs2DUFBQXB3d0dUVBROnDhRO41tQG71uYwePbrCd2jgwIG109gGZObMmbj33nvh5eWFgIAAxMTE4Pjx4w41hYWFGDduHBo3boxGjRph6NChFW6w3pAxGNUBK1euxKRJk5CQkIB9+/aha9euiI6ORmZmZm03rUHr2LEjLl26pEw7duyo7SY1SHl5eejatSvmz59f6fLZs2dj3rx5WLRoEVJSUuDp6Yno6GgUFhbe4ZY2LLf6XABg4MCBDt+h5cuX38EWNkw///wzxo0bh927dyMxMRElJSV4+OGHkZeXp9S88sorWLduHVavXo2ff/4ZFy9exJ/+9KdabHUdU+W7p1KN6dmzpxg3bpzy3GazieDgYDFz5sxabFXDlpCQILp27VrbzaDrABDffvut8txutwuj0SjmzJmjzDObzUKr1Yrly5fXQgsbpus/FyGEGDVqlBgyZEittIeuyczMFADEzz//LISQvh9ubm5i9erVSs2xY8cEAJGcnFxbzaxT2GNUy4qLi5GamoqoqChlnlqtRlRUFJKTk2uxZXTixAkEBwejRYsWGDFiBM6ePVvbTaLrpKenw2QyOXx/9Ho9IiIi+P2pA7Zu3YqAgAC0bdsWL7zwAq5cuVLbTWpwLBYLAMDX1xcAkJqaipKSEofvTLt27dCsWTN+Z2QMRrXs8uXLsNlsCAwMdJgfGBgIk8lUS62iiIgILF26FJs2bcLChQuRnp6OPn36ICcnp7abRuWUfUf4/al7Bg4ciP/9739ISkrCu+++i59//hmPPPIIbDZbbTetwbDb7Zg4cSJ69eqFTp06AZC+MxqNBgaDwaGW35lrXGu7AUR10SOPPKL8uUuXLoiIiEDz5s2xatUqxMXF1WLLiO4OTz75pPLnzp07o0uXLmjZsiW2bt2KAQMG1GLLGo5x48bhyJEjHB/pJPYY1TI/Pz+4uLhUOCMgIyMDRqOxllpF1zMYDGjTpg1OnjxZ202hcsq+I/z+1H0tWrSAn58fv0N3yPjx47F+/Xps2bIFTZs2VeYbjUYUFxfDbDY71PM7cw2DUS3TaDTo3r07kpKSlHl2ux1JSUmIjIysxZZRebm5uTh16hSCgoJquylUTlhYGIxGo8P3x2q1IiUlhd+fOub8+fO4cuUKv0M1TAiB8ePH49tvv8XmzZsRFhbmsLx79+5wc3Nz+M4cP34cZ8+e5XdGxkNpdcCkSZMwatQo9OjRAz179sQHH3yAvLw8jBkzprab1mC9+uqrGDx4MJo3b46LFy8iISEBLi4uiI2Nre2mNTi5ubkOvQzp6ek4cOAAfH190axZM0ycOBFvv/02WrdujbCwMEydOhXBwcGIiYmpvUY3ADf7XHx9fTFjxgwMHToURqMRp06dwmuvvYZWrVohOjq6Fltd/40bNw5ffvkl1q5dCy8vL2XckF6vh7u7O/R6PeLi4jBp0iT4+vrC29sbL730EiIjI3HffffVcuvriNo+LY4kH374oWjWrJnQaDSiZ8+eYvfu3bXdpAZt+PDhIigoSGg0GtGkSRMxfPhwcfLkydpuVoO0ZcsWAaDCNGrUKCGEdMr+1KlTRWBgoNBqtWLAgAHi+PHjtdvoBuBmn0t+fr54+OGHhb+/v3BzcxPNmzcXzz33nDCZTLXd7Hqvss8EgFiyZIlSU1BQIF588UXh4+MjPDw8xBNPPCEuXbpUe42uY1RCCHHn4xgRERFR3cMxRkREREQyBiMiIiIiGYMRERERkYzBiIiIiEjGYEREREQkYzAiIiIikjEYEREREckYjIiIiIhkDEZEREREMgYjIiIiIhmDEREREZGMwYiIiIhI9v/uB2EoH1rgIQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# LSTM\n",
    "plot_predictions(inv_y_lstm, inv_yhat_lstm, model_name = 'LSTM')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1edacf5d",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:generalml_p37_gpu_v1]",
   "language": "python",
   "name": "conda-env-generalml_p37_gpu_v1-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}