Report 2 - DATA CLEANING.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Data cleaning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:12:48.037644Z",
     "start_time": "2020-03-01T09:12:43.385755Z"
    }
   },
   "outputs": [],
   "source": [
    "import pandas as pd \n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from statistics import mode\n",
    "from keras.utils import to_categorical\n",
    "from sklearn import preprocessing\n",
    "from sklearn.model_selection import GridSearchCV, RandomizedSearchCV\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "from tensorflow.python.keras.layers import Dense\n",
    "from tensorflow.python.keras import Sequential\n",
    "from tensorflow.python.keras.layers import Dropout\n",
    "from tensorflow.python.keras.initializers import he_normal, glorot_normal\n",
    "from tensorflow.keras.optimizers import Adam, SGD\n",
    "from tensorflow.keras.regularizers import l1,l2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.1 Preliminary steps"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.1.1 Uploading the dataset and removing the duplicates."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:12:48.136583Z",
     "start_time": "2020-03-01T09:12:48.051634Z"
    }
   },
   "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>ID</th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>...</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>34579</td>\n",
       "      <td>35</td>\n",
       "      <td>admin.</td>\n",
       "      <td>single</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>may</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>1</td>\n",
       "      <td>failure</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.266</td>\n",
       "      <td>5099.1</td>\n",
       "      <td>no</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>446</td>\n",
       "      <td>42</td>\n",
       "      <td>technician</td>\n",
       "      <td>married</td>\n",
       "      <td>professional.course</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "      <td>yes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>20173</td>\n",
       "      <td>36</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>aug</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.444</td>\n",
       "      <td>-36.1</td>\n",
       "      <td>4.965</td>\n",
       "      <td>5228.1</td>\n",
       "      <td>yes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>18171</td>\n",
       "      <td>37</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>yes</td>\n",
       "      <td>telephone</td>\n",
       "      <td>jul</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.918</td>\n",
       "      <td>-42.7</td>\n",
       "      <td>4.963</td>\n",
       "      <td>5228.1</td>\n",
       "      <td>yes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>30128</td>\n",
       "      <td>31</td>\n",
       "      <td>management</td>\n",
       "      <td>single</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>apr</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>93.075</td>\n",
       "      <td>-47.1</td>\n",
       "      <td>1.365</td>\n",
       "      <td>5099.1</td>\n",
       "      <td>no</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      ID  age         job  marital            education default housing loan  \\\n",
       "0  34579   35      admin.   single    university.degree      no     yes   no   \n",
       "1    446   42  technician  married  professional.course      no      no   no   \n",
       "2  20173   36      admin.  married    university.degree      no      no   no   \n",
       "3  18171   37      admin.  married          high.school      no     yes  yes   \n",
       "4  30128   31  management   single    university.degree      no     yes   no   \n",
       "\n",
       "     contact month  ... campaign  pdays  previous     poutcome emp.var.rate  \\\n",
       "0   cellular   may  ...        1    999         1      failure         -1.8   \n",
       "1  telephone   may  ...        1    999         0  nonexistent          1.1   \n",
       "2   cellular   aug  ...        2    999         0  nonexistent          1.4   \n",
       "3  telephone   jul  ...        2    999         0  nonexistent          1.4   \n",
       "4   cellular   apr  ...        1    999         0  nonexistent         -1.8   \n",
       "\n",
       "   cons.price.idx  cons.conf.idx  euribor3m  nr.employed    y  \n",
       "0          92.893          -46.2      1.266       5099.1   no  \n",
       "1          93.994          -36.4      4.857       5191.0  yes  \n",
       "2          93.444          -36.1      4.965       5228.1  yes  \n",
       "3          93.918          -42.7      4.963       5228.1  yes  \n",
       "4          93.075          -47.1      1.365       5099.1   no  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv('BankMarketing.csv',sep=';')\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:12:48.238518Z",
     "start_time": "2020-03-01T09:12:48.200543Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "New shape of the dataset: 8980 20\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>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ID</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>34579</th>\n",
       "      <td>35</td>\n",
       "      <td>admin.</td>\n",
       "      <td>single</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>may</td>\n",
       "      <td>thu</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>1</td>\n",
       "      <td>failure</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.266</td>\n",
       "      <td>5099.1</td>\n",
       "      <td>no</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>446</th>\n",
       "      <td>42</td>\n",
       "      <td>technician</td>\n",
       "      <td>married</td>\n",
       "      <td>professional.course</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>tue</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "      <td>yes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20173</th>\n",
       "      <td>36</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>aug</td>\n",
       "      <td>mon</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.444</td>\n",
       "      <td>-36.1</td>\n",
       "      <td>4.965</td>\n",
       "      <td>5228.1</td>\n",
       "      <td>yes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18171</th>\n",
       "      <td>37</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>yes</td>\n",
       "      <td>telephone</td>\n",
       "      <td>jul</td>\n",
       "      <td>wed</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.918</td>\n",
       "      <td>-42.7</td>\n",
       "      <td>4.963</td>\n",
       "      <td>5228.1</td>\n",
       "      <td>yes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30128</th>\n",
       "      <td>31</td>\n",
       "      <td>management</td>\n",
       "      <td>single</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>apr</td>\n",
       "      <td>thu</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>93.075</td>\n",
       "      <td>-47.1</td>\n",
       "      <td>1.365</td>\n",
       "      <td>5099.1</td>\n",
       "      <td>no</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       age         job  marital            education default housing loan  \\\n",
       "ID                                                                          \n",
       "34579   35      admin.   single    university.degree      no     yes   no   \n",
       "446     42  technician  married  professional.course      no      no   no   \n",
       "20173   36      admin.  married    university.degree      no      no   no   \n",
       "18171   37      admin.  married          high.school      no     yes  yes   \n",
       "30128   31  management   single    university.degree      no     yes   no   \n",
       "\n",
       "         contact month day_of_week  campaign  pdays  previous     poutcome  \\\n",
       "ID                                                                           \n",
       "34579   cellular   may         thu         1    999         1      failure   \n",
       "446    telephone   may         tue         1    999         0  nonexistent   \n",
       "20173   cellular   aug         mon         2    999         0  nonexistent   \n",
       "18171  telephone   jul         wed         2    999         0  nonexistent   \n",
       "30128   cellular   apr         thu         1    999         0  nonexistent   \n",
       "\n",
       "       emp.var.rate  cons.price.idx  cons.conf.idx  euribor3m  nr.employed  \\\n",
       "ID                                                                           \n",
       "34579          -1.8          92.893          -46.2      1.266       5099.1   \n",
       "446             1.1          93.994          -36.4      4.857       5191.0   \n",
       "20173           1.4          93.444          -36.1      4.965       5228.1   \n",
       "18171           1.4          93.918          -42.7      4.963       5228.1   \n",
       "30128          -1.8          93.075          -47.1      1.365       5099.1   \n",
       "\n",
       "         y  \n",
       "ID          \n",
       "34579   no  \n",
       "446    yes  \n",
       "20173  yes  \n",
       "18171  yes  \n",
       "30128   no  "
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.drop_duplicates(inplace=True)\n",
    "data.set_index('ID',inplace=True)\n",
    "print('New shape of the dataset:',*data.shape)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.1.2 Checking the correct reading of the variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:12:49.145350Z",
     "start_time": "2020-03-01T09:12:49.135357Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "age                 int64\n",
       "job                object\n",
       "marital            object\n",
       "education          object\n",
       "default            object\n",
       "housing            object\n",
       "loan               object\n",
       "contact            object\n",
       "month              object\n",
       "day_of_week        object\n",
       "campaign            int64\n",
       "pdays               int64\n",
       "previous            int64\n",
       "poutcome           object\n",
       "emp.var.rate      float64\n",
       "cons.price.idx    float64\n",
       "cons.conf.idx     float64\n",
       "euribor3m         float64\n",
       "nr.employed       float64\n",
       "y                  object\n",
       "dtype: object"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.dtypes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Categorical variables are read as `object` instead of `category`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:12:50.915502Z",
     "start_time": "2020-03-01T09:12:50.878525Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "age                  int64\n",
       "job               category\n",
       "marital           category\n",
       "education         category\n",
       "default           category\n",
       "housing           category\n",
       "loan              category\n",
       "contact           category\n",
       "month             category\n",
       "day_of_week       category\n",
       "campaign             int64\n",
       "pdays                int64\n",
       "previous             int64\n",
       "poutcome          category\n",
       "emp.var.rate       float64\n",
       "cons.price.idx     float64\n",
       "cons.conf.idx      float64\n",
       "euribor3m          float64\n",
       "nr.employed        float64\n",
       "y                 category\n",
       "dtype: object"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cat=list(dict(data.dtypes[data.dtypes=='object']).keys())\n",
    "\n",
    "for i in cat: \n",
    "    data[i]=data[i].astype('category')\n",
    "\n",
    "data.dtypes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.2 Reparameterizing `pdays`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:12:52.547619Z",
     "start_time": "2020-03-01T09:12:52.533630Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    8980.000000\n",
       "mean      885.204232\n",
       "std       316.342109\n",
       "min         0.000000\n",
       "25%       999.000000\n",
       "50%       999.000000\n",
       "75%       999.000000\n",
       "max       999.000000\n",
       "Name: pdays, dtype: float64"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['pdays'].describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`pdays` is currently being read as a continuous variable. Unfortunately there is the persistence of a value that is out of scale with respect to the others: `999`. We suppose that this value stands for a number of days that goes beyond the third year. For this reason, we suppose that people having this value have not been contacted recently. \n",
    "\n",
    "We have decided to reparametrize the variable into a categorical one with 4 levels: \n",
    "- `1_w`: people that have been contacted \"within a week\"\n",
    "- `2_w`: people that have been contacted \"within two weeks\"\n",
    "- `within_month`: people that have been contacted \"within the month\"\n",
    "- `not_rec_cont`: which stands for people that have not been recently contacted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:12:53.703617Z",
     "start_time": "2020-03-01T09:12:53.678631Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "not_rec_cont    7951\n",
      "1_w              821\n",
      "2_w              168\n",
      "within_month      40\n",
      "Name: pdays_new, dtype: int64\n",
      "category\n"
     ]
    }
   ],
   "source": [
    "l=list(data['pdays'])\n",
    "\n",
    "pdays_new=list()\n",
    "for i in l:\n",
    "    if i >=0 and i<=7:\n",
    "        pdays_new.append('1_w')\n",
    "    elif i>=8 and i<=14:\n",
    "        pdays_new.append('2_w')\n",
    "    elif i>=15 and i<=31:\n",
    "        pdays_new.append('within_month')\n",
    "    elif i==999:\n",
    "        pdays_new.append('not_rec_cont')\n",
    "        \n",
    "data['pdays_new']=pdays_new\n",
    "data['pdays_new']=data['pdays_new'].astype('category')\n",
    "data.drop('pdays',axis=1,inplace=True)\n",
    "\n",
    "print(data['pdays_new'].value_counts())\n",
    "print(data['pdays_new'].dtypes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:11:04.783947Z",
     "start_time": "2020-03-01T09:11:04.116358Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "          index  Perc.%\n",
      "0  not_rec_cont   88.54\n",
      "1           1_w    9.14\n",
      "2           2_w    1.87\n",
      "3  within_month    0.45\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x576 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set_style('whitegrid')\n",
    "sns.set(font_scale = 1.25)\n",
    "\n",
    "pd = (data['pdays_new'].value_counts(normalize=True)\n",
    "                     .rename('Perc.%')\n",
    "                     .mul(100)\n",
    "                     .reset_index())\n",
    "print(round(pd,2))\n",
    "\n",
    "fig, (ax1, ax2) = plt.subplots(nrows = 1, ncols = 2, figsize = (20,8))\n",
    "\n",
    "a=sns.barplot(x='index',y='Perc.%',data=pd,ax=ax1)\n",
    "a.set(xlabel='Pdays')\n",
    "\n",
    "\n",
    "pd_y = (data.groupby(['pdays_new'])['y']\n",
    "         .value_counts(normalize=True)\n",
    "         .rename('Perc.%')\n",
    "         .mul(100)\n",
    "         .reset_index())\n",
    "\n",
    "sns.barplot(x=\"pdays_new\", y=\"Perc.%\", hue=\"y\", data=pd_y,ax=ax2)\n",
    "plt.xlabel('Pdays')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.3 `unknown` imputation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "During the descriptive analysis, we noticed the anomalous presence of `unknown` values in multiple variables. We suspect that these values are missing, so we decided to find ways to compute (impute) these missing values."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.3.1 Checking the allocation of all the `unknown` attributes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The first thing we wanted to do is to check is the quantity of `unknown` values that are present in each variable."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:12:56.362613Z",
     "start_time": "2020-03-01T09:12:56.313642Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In absolute terms:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'age': 0,\n",
       " 'job': 75,\n",
       " 'marital': 21,\n",
       " 'education': 425,\n",
       " 'default': 1443,\n",
       " 'housing': 209,\n",
       " 'loan': 209,\n",
       " 'contact': 0,\n",
       " 'month': 0,\n",
       " 'day_of_week': 0,\n",
       " 'campaign': 0,\n",
       " 'previous': 0,\n",
       " 'poutcome': 0,\n",
       " 'emp.var.rate': 0,\n",
       " 'cons.price.idx': 0,\n",
       " 'cons.conf.idx': 0,\n",
       " 'euribor3m': 0,\n",
       " 'nr.employed': 0,\n",
       " 'y': 0,\n",
       " 'pdays_new': 0}"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unk_counter=dict()\n",
    "\n",
    "for i in data.columns:\n",
    "    unk_counter[i]=sum(data.loc[:,i]=='unknown')\n",
    "\n",
    "print('In absolute terms:')\n",
    "unk_counter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:01.144805Z",
     "start_time": "2020-03-01T09:13:01.093836Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In frequency terms:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'age': 0.0,\n",
       " 'job': 0.0084,\n",
       " 'marital': 0.0023,\n",
       " 'education': 0.0473,\n",
       " 'default': 0.1607,\n",
       " 'housing': 0.0233,\n",
       " 'loan': 0.0233,\n",
       " 'contact': 0.0,\n",
       " 'month': 0.0,\n",
       " 'day_of_week': 0.0,\n",
       " 'campaign': 0.0,\n",
       " 'previous': 0.0,\n",
       " 'poutcome': 0.0,\n",
       " 'emp.var.rate': 0.0,\n",
       " 'cons.price.idx': 0.0,\n",
       " 'cons.conf.idx': 0.0,\n",
       " 'euribor3m': 0.0,\n",
       " 'nr.employed': 0.0,\n",
       " 'y': 0.0,\n",
       " 'pdays_new': 0.0}"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unk_freq=dict()\n",
    "\n",
    "for i in data.columns:\n",
    "    unk_freq[i]=round(sum(data.loc[:,i]=='unknown')/len(data),4)\n",
    "    \n",
    "print('In frequency terms:')\n",
    "unk_freq"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It is possible to see that the all the `unknown` values are present in the client personal information part of the dataset. Especially in the `default` variable. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.3.2 Default"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From the descriptive analysis it is possible to see that the only two levels of `default` are `No` and `unknown`. For this reason it is impossible to deduce if these `unknown` are considered `Yes` or `No`, and due that, we decided to drop this variable from the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:04.600006Z",
     "start_time": "2020-03-01T09:13:04.592009Z"
    }
   },
   "outputs": [],
   "source": [
    "data.drop('default',axis=1,inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.3.3 Job and marital "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The amount of missing values for these two variables is very low. For job they just correspond to the 0.84% of the total amount of the variable itself, while for marital just the 0.23%.<br>\n",
    "Due this, we substitute the missing values in both of the variables with the mode."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:06.237514Z",
     "start_time": "2020-03-01T09:13:06.214530Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Job:\n",
      "admin.           2555\n",
      "blue-collar      1663\n",
      "technician       1430\n",
      "services          756\n",
      "management        633\n",
      "retired           569\n",
      "student           326\n",
      "self-employed     306\n",
      "entrepreneur      281\n",
      "unemployed        235\n",
      "housemaid         226\n",
      "Name: job, dtype: int64\n",
      "\n",
      "Marital:\n",
      "married     5236\n",
      "single      2791\n",
      "divorced     953\n",
      "Name: marital, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "for i in data[['job','marital']]:\n",
    "    data[i].replace({'unknown':mode(data[i])},inplace=True)\n",
    "print('Job:')\n",
    "print(data['job'].value_counts())\n",
    "print('')\n",
    "print('Marital:')\n",
    "print(data['marital'].value_counts())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.4 Education imputation through non-parametric approaches "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Education housing and loan are the variables with the higher amount of missing values. We will compute them with two different non-parametric approaches: Random forest and ANN. Both of them will be trained on the subset of data with the present values and compared using the crossvalidation approach on the subset. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.4.1 One hot encoding & train, test and validation splitting"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before training the algorithms, we need to encode the dataset by doing one-hot-encoding since in the dataset there are a lot of non categorical variables.<br>\n",
    "Moreover we normalize the numerical variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_origin=data.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:08.773466Z",
     "start_time": "2020-03-01T09:13:08.704511Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The columns are now the following:\n",
      "\n",
      "Index(['age', 'campaign', 'cons.conf.idx', 'cons.price.idx', 'emp.var.rate',\n",
      "       'euribor3m', 'nr.employed', 'previous', 'day_of_week_fri',\n",
      "       'day_of_week_mon', 'day_of_week_thu', 'day_of_week_tue',\n",
      "       'day_of_week_wed', 'job_admin.', 'job_blue-collar', 'job_entrepreneur',\n",
      "       'job_housemaid', 'job_management', 'job_retired', 'job_self-employed',\n",
      "       'job_services', 'job_student', 'job_technician', 'job_unemployed',\n",
      "       'marital_divorced', 'marital_married', 'marital_single', 'month_apr',\n",
      "       'month_aug', 'month_dec', 'month_jul', 'month_jun', 'month_mar',\n",
      "       'month_may', 'month_nov', 'month_oct', 'month_sep', 'pdays_new_1_w',\n",
      "       'pdays_new_2_w', 'pdays_new_not_rec_cont', 'pdays_new_within_month',\n",
      "       'poutcome_failure', 'poutcome_nonexistent', 'poutcome_success'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "numi=['age','campaign','cons.conf.idx','cons.price.idx','emp.var.rate','euribor3m','nr.employed','previous']\n",
    "data_origin[numi]=preprocessing.normalize(data_origin[numi])\n",
    "\n",
    "data_edu=data_origin[data_origin.columns.difference(['loan','housing','education','y','contact'])]\n",
    "    \n",
    "data_one=pd.get_dummies(data_edu)\n",
    "\n",
    "print('The columns are now the following:')\n",
    "print('')\n",
    "print(data_one.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:14.942343Z",
     "start_time": "2020-03-01T09:13:14.912362Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "age                        float64\n",
       "campaign                   float64\n",
       "cons.conf.idx              float64\n",
       "cons.price.idx             float64\n",
       "emp.var.rate               float64\n",
       "euribor3m                  float64\n",
       "nr.employed                float64\n",
       "previous                   float64\n",
       "day_of_week_fri              uint8\n",
       "day_of_week_mon              uint8\n",
       "day_of_week_thu              uint8\n",
       "day_of_week_tue              uint8\n",
       "day_of_week_wed              uint8\n",
       "job_admin.                   uint8\n",
       "job_blue-collar              uint8\n",
       "job_entrepreneur             uint8\n",
       "job_housemaid                uint8\n",
       "job_management               uint8\n",
       "job_retired                  uint8\n",
       "job_self-employed            uint8\n",
       "job_services                 uint8\n",
       "job_student                  uint8\n",
       "job_technician               uint8\n",
       "job_unemployed               uint8\n",
       "marital_divorced             uint8\n",
       "marital_married              uint8\n",
       "marital_single               uint8\n",
       "month_apr                    uint8\n",
       "month_aug                    uint8\n",
       "month_dec                    uint8\n",
       "month_jul                    uint8\n",
       "month_jun                    uint8\n",
       "month_mar                    uint8\n",
       "month_may                    uint8\n",
       "month_nov                    uint8\n",
       "month_oct                    uint8\n",
       "month_sep                    uint8\n",
       "pdays_new_1_w                uint8\n",
       "pdays_new_2_w                uint8\n",
       "pdays_new_not_rec_cont       uint8\n",
       "pdays_new_within_month       uint8\n",
       "poutcome_failure             uint8\n",
       "poutcome_nonexistent         uint8\n",
       "poutcome_success             uint8\n",
       "education                 category\n",
       "contact                       int8\n",
       "y                            int64\n",
       "dtype: object"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_one=data_one.merge(data[['education','contact','y']],on='ID')\n",
    "\n",
    "data_one['y'].replace({'yes':1,'no':0},inplace=True)\n",
    "\n",
    "\n",
    "data_one['contact']=data_one['contact'].astype('category')\n",
    "data_one['contact']=data_one['contact'].cat.codes\n",
    "data_one.dtypes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:29.453584Z",
     "start_time": "2020-03-01T09:13:29.422604Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The shape of the dataset now is the following:\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>age</th>\n",
       "      <th>campaign</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "      <th>previous</th>\n",
       "      <th>day_of_week_fri</th>\n",
       "      <th>day_of_week_mon</th>\n",
       "      <th>...</th>\n",
       "      <th>pdays_new_1_w</th>\n",
       "      <th>pdays_new_2_w</th>\n",
       "      <th>pdays_new_not_rec_cont</th>\n",
       "      <th>pdays_new_within_month</th>\n",
       "      <th>poutcome_failure</th>\n",
       "      <th>poutcome_nonexistent</th>\n",
       "      <th>poutcome_success</th>\n",
       "      <th>education</th>\n",
       "      <th>contact</th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ID</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>34579</th>\n",
       "      <td>0.006862</td>\n",
       "      <td>0.000196</td>\n",
       "      <td>-0.009058</td>\n",
       "      <td>0.018213</td>\n",
       "      <td>-0.000353</td>\n",
       "      <td>0.000248</td>\n",
       "      <td>0.999769</td>\n",
       "      <td>0.000196</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>446</th>\n",
       "      <td>0.008089</td>\n",
       "      <td>0.000193</td>\n",
       "      <td>-0.007011</td>\n",
       "      <td>0.018103</td>\n",
       "      <td>0.000212</td>\n",
       "      <td>0.000935</td>\n",
       "      <td>0.999778</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>professional.course</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20173</th>\n",
       "      <td>0.006884</td>\n",
       "      <td>0.000382</td>\n",
       "      <td>-0.006904</td>\n",
       "      <td>0.017870</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>0.000949</td>\n",
       "      <td>0.999792</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18171</th>\n",
       "      <td>0.007076</td>\n",
       "      <td>0.000382</td>\n",
       "      <td>-0.008166</td>\n",
       "      <td>0.017960</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>0.000949</td>\n",
       "      <td>0.999780</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>high.school</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30128</th>\n",
       "      <td>0.006078</td>\n",
       "      <td>0.000196</td>\n",
       "      <td>-0.009235</td>\n",
       "      <td>0.018249</td>\n",
       "      <td>-0.000353</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>0.999772</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 47 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            age  campaign  cons.conf.idx  cons.price.idx  emp.var.rate  \\\n",
       "ID                                                                       \n",
       "34579  0.006862  0.000196      -0.009058        0.018213     -0.000353   \n",
       "446    0.008089  0.000193      -0.007011        0.018103      0.000212   \n",
       "20173  0.006884  0.000382      -0.006904        0.017870      0.000268   \n",
       "18171  0.007076  0.000382      -0.008166        0.017960      0.000268   \n",
       "30128  0.006078  0.000196      -0.009235        0.018249     -0.000353   \n",
       "\n",
       "       euribor3m  nr.employed  previous  day_of_week_fri  day_of_week_mon  \\\n",
       "ID                                                                          \n",
       "34579   0.000248     0.999769  0.000196                0                0   \n",
       "446     0.000935     0.999778  0.000000                0                0   \n",
       "20173   0.000949     0.999792  0.000000                0                1   \n",
       "18171   0.000949     0.999780  0.000000                0                0   \n",
       "30128   0.000268     0.999772  0.000000                0                0   \n",
       "\n",
       "       ...  pdays_new_1_w  pdays_new_2_w  pdays_new_not_rec_cont  \\\n",
       "ID     ...                                                         \n",
       "34579  ...              0              0                       1   \n",
       "446    ...              0              0                       1   \n",
       "20173  ...              0              0                       1   \n",
       "18171  ...              0              0                       1   \n",
       "30128  ...              0              0                       1   \n",
       "\n",
       "       pdays_new_within_month  poutcome_failure  poutcome_nonexistent  \\\n",
       "ID                                                                      \n",
       "34579                       0                 1                     0   \n",
       "446                         0                 0                     1   \n",
       "20173                       0                 0                     1   \n",
       "18171                       0                 0                     1   \n",
       "30128                       0                 0                     1   \n",
       "\n",
       "       poutcome_success            education  contact  y  \n",
       "ID                                                        \n",
       "34579                 0    university.degree        0  0  \n",
       "446                   0  professional.course        1  1  \n",
       "20173                 0    university.degree        0  1  \n",
       "18171                 0          high.school        1  1  \n",
       "30128                 0    university.degree        0  0  \n",
       "\n",
       "[5 rows x 47 columns]"
      ]
     },
     "execution_count": 90,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('The shape of the dataset now is the following:')\n",
    "data_one.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that the whole dataset has been reparametrized, we divide the dataset into a train set (`sub_ed`) where the algorithms will be compared and a validation set (`val_ed`) where the `unknown` values are contained. On this last subset, predictions will be made and the `unknown` attributes will be substituted."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:37.861222Z",
     "start_time": "2020-03-01T09:13:37.846234Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of the train set: (8555, 47)\n",
      "\n",
      "Shape of the validation set: (425, 47)\n"
     ]
    }
   ],
   "source": [
    "val_ed=data_one[data_one['education']=='unknown']\n",
    "sub_ed=data_one[data_one['education']!='unknown']\n",
    "\n",
    "print('Shape of the train set:',sub_ed.shape)\n",
    "print('')\n",
    "print('Shape of the validation set:',val_ed.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From the subset `sub_ed` we label encode the variable `education` since now we don't have anymore the `unknown` parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:41.069721Z",
     "start_time": "2020-03-01T09:13:40.958789Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "ID\n",
      "34579    6\n",
      "446      5\n",
      "20173    6\n",
      "18171    4\n",
      "30128    6\n",
      "Name: education, dtype: int64\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ANDREA\\Anaconda3\\lib\\site-packages\\pandas\\core\\generic.py:6786: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  self._update_inplace(new_data)\n"
     ]
    }
   ],
   "source": [
    "print('')\n",
    "di = {'illiterate': 0, 'professional.course':5, 'high.school':4,'university.degree':6, 'basic.9y':3, 'basic.4y':1, 'basic.6y':2 }\n",
    "sub_ed[\"education\"].replace(di, inplace=True)\n",
    "print(sub_ed['education'].head())\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We now can explicit the X (the data) and the Y (the response) for helping us to make thing more easy further.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:13:44.608184Z",
     "start_time": "2020-03-01T09:13:44.599190Z"
    }
   },
   "outputs": [],
   "source": [
    "X=sub_ed[sub_ed.columns.difference(['education'])]\n",
    "Y=sub_ed['education']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.4.2 First model: Random Forest"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For the random forest we decided to perform a k=5 fold cross-validation for estimating the best criterion, the maximum depth that each tree should have, the minimum samples leaf and the max of the features. Moreover, we did this by crossvalidating for 5 different forests, each one with an increasing number of trees to be estimated."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:19:29.487513Z",
     "start_time": "2020-03-01T09:13:48.789894Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "50   0.5263588544710696   {'min_samples_leaf': 0.001, 'max_features': 0.2, 'max_depth': 5, 'criterion': 'entropy'}\n",
      "100   0.5241379310344828   {'min_samples_leaf': 0.001, 'max_features': 0.2, 'max_depth': 5, 'criterion': 'entropy'}\n",
      "150   0.520864991233197   {'min_samples_leaf': 0.001, 'max_features': 0.2, 'max_depth': 5, 'criterion': 'entropy'}\n",
      "200   0.5226183518410287   {'min_samples_leaf': 0.001, 'max_features': 0.2, 'max_depth': 5, 'criterion': 'entropy'}\n"
     ]
    }
   ],
   "source": [
    "for i in [50,100,150,200]:\n",
    "    rf = RandomForestClassifier(random_state=33, n_estimators=i)\n",
    "    params = {'criterion': ['gini', 'entropy'], 'max_depth': np.arange(2, 6), 'min_samples_leaf': np.logspace(-1, -3, 3),'max_features': ['sqrt', 0.2, 0.5]}\n",
    "    cv_tuning = RandomizedSearchCV(rf, params, random_state=122223, scoring='accuracy', return_train_score=True, cv=5, n_iter=20)\n",
    "    cv_tuning.fit(X,Y)\n",
    "    print(i,' ',cv_tuning.best_score_,' ',cv_tuning.best_params_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-02-29T17:39:02.380880Z",
     "start_time": "2020-02-29T17:39:02.109539Z"
    }
   },
   "source": [
    "As it is possible to see, the best random forest is the first one with 50 trees which has an accuracy of 52,6%. The algorithm in this form is more computationally heavy, but it allows us to see the best model for each different forest."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.4.3 Second model: Artificial Neural Network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here, we need to perform a train-test split since a crossvalidation for evaluating the best parameters is too computationally costly. Also, for this reason, we will test different configuration in order to find the one that suits better the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:19:29.803512Z",
     "start_time": "2020-03-01T09:19:29.764344Z"
    }
   },
   "outputs": [],
   "source": [
    "train_x,test_x,train_y,test_y=train_test_split(X, Y, test_size=0.2, random_state=123, stratify=Y,shuffle=True)\n",
    "\n",
    "train_y_one=pd.get_dummies(train_y)\n",
    "test_y_one=pd.get_dummies(test_y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Config. 1: high batch size  and small learning rate in order to stabilize the training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:19:38.880526Z",
     "start_time": "2020-03-01T09:19:30.096330Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Falling back from v2 loop because of error: Failed to find data adapter that can handle input: <class 'pandas.core.frame.DataFrame'>, <class 'NoneType'>\n",
      "Train on 6844 samples, validate on 1711 samples\n",
      "Epoch 1/150\n",
      "6844/6844 [==============================] - 1s 125us/sample - loss: 1.9081 - accuracy: 0.2183 - val_loss: 1.8629 - val_accuracy: 0.2811\n",
      "Epoch 2/150\n",
      "6844/6844 [==============================] - 0s 14us/sample - loss: 1.8261 - accuracy: 0.3074 - val_loss: 1.7802 - val_accuracy: 0.3302\n",
      "Epoch 3/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.7425 - accuracy: 0.3343 - val_loss: 1.6897 - val_accuracy: 0.3477\n",
      "Epoch 4/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.6551 - accuracy: 0.3413 - val_loss: 1.6054 - val_accuracy: 0.3437\n",
      "Epoch 5/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.5787 - accuracy: 0.3460 - val_loss: 1.5365 - val_accuracy: 0.3565\n",
      "Epoch 6/150\n",
      "6844/6844 [==============================] - ETA: 0s - loss: 1.5210 - accuracy: 0.36 - 0s 11us/sample - loss: 1.5158 - accuracy: 0.3717 - val_loss: 1.4780 - val_accuracy: 0.3805\n",
      "Epoch 7/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.4615 - accuracy: 0.3926 - val_loss: 1.4241 - val_accuracy: 0.4021\n",
      "Epoch 8/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.4111 - accuracy: 0.4147 - val_loss: 1.3756 - val_accuracy: 0.4237\n",
      "Epoch 9/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.3658 - accuracy: 0.4500 - val_loss: 1.3323 - val_accuracy: 0.4681\n",
      "Epoch 10/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.3247 - accuracy: 0.4896 - val_loss: 1.2954 - val_accuracy: 0.5032\n",
      "Epoch 11/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.2892 - accuracy: 0.5134 - val_loss: 1.2670 - val_accuracy: 0.5143\n",
      "Epoch 12/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.2621 - accuracy: 0.5273 - val_loss: 1.2459 - val_accuracy: 0.5289\n",
      "Epoch 13/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.2419 - accuracy: 0.5346 - val_loss: 1.2323 - val_accuracy: 0.5272\n",
      "Epoch 14/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.2277 - accuracy: 0.5376 - val_loss: 1.2240 - val_accuracy: 0.5295\n",
      "Epoch 15/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.2199 - accuracy: 0.5373 - val_loss: 1.2182 - val_accuracy: 0.5371\n",
      "Epoch 16/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.2117 - accuracy: 0.5390 - val_loss: 1.2147 - val_accuracy: 0.5354\n",
      "Epoch 17/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.2073 - accuracy: 0.5367 - val_loss: 1.2130 - val_accuracy: 0.5295\n",
      "Epoch 18/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.2030 - accuracy: 0.5390 - val_loss: 1.2112 - val_accuracy: 0.5313\n",
      "Epoch 19/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.2005 - accuracy: 0.5381 - val_loss: 1.2114 - val_accuracy: 0.5336\n",
      "Epoch 20/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1975 - accuracy: 0.5406 - val_loss: 1.2091 - val_accuracy: 0.5319\n",
      "Epoch 21/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1953 - accuracy: 0.5408 - val_loss: 1.2076 - val_accuracy: 0.5324\n",
      "Epoch 22/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1930 - accuracy: 0.5399 - val_loss: 1.2093 - val_accuracy: 0.5301\n",
      "Epoch 23/150\n",
      "6844/6844 [==============================] - 0s 13us/sample - loss: 1.1914 - accuracy: 0.5411 - val_loss: 1.2067 - val_accuracy: 0.5283\n",
      "Epoch 24/150\n",
      "6844/6844 [==============================] - 0s 15us/sample - loss: 1.1900 - accuracy: 0.5395 - val_loss: 1.2064 - val_accuracy: 0.5324\n",
      "Epoch 25/150\n",
      "6844/6844 [==============================] - 0s 15us/sample - loss: 1.1904 - accuracy: 0.5425 - val_loss: 1.2093 - val_accuracy: 0.5365\n",
      "Epoch 26/150\n",
      "6844/6844 [==============================] - 0s 15us/sample - loss: 1.1877 - accuracy: 0.5406 - val_loss: 1.2051 - val_accuracy: 0.5295\n",
      "Epoch 27/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1864 - accuracy: 0.5419 - val_loss: 1.2051 - val_accuracy: 0.5319\n",
      "Epoch 28/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1845 - accuracy: 0.5440 - val_loss: 1.2049 - val_accuracy: 0.5365\n",
      "Epoch 29/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1837 - accuracy: 0.5430 - val_loss: 1.2044 - val_accuracy: 0.5342\n",
      "Epoch 30/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1834 - accuracy: 0.5428 - val_loss: 1.2046 - val_accuracy: 0.5295\n",
      "Epoch 31/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1811 - accuracy: 0.5430 - val_loss: 1.2048 - val_accuracy: 0.5395\n",
      "Epoch 32/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1804 - accuracy: 0.5424 - val_loss: 1.2041 - val_accuracy: 0.5359\n",
      "Epoch 33/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1791 - accuracy: 0.5432 - val_loss: 1.2045 - val_accuracy: 0.5278\n",
      "Epoch 34/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1782 - accuracy: 0.5446 - val_loss: 1.2032 - val_accuracy: 0.5330\n",
      "Epoch 35/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1766 - accuracy: 0.5422 - val_loss: 1.2042 - val_accuracy: 0.5354\n",
      "Epoch 36/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1765 - accuracy: 0.5450 - val_loss: 1.2028 - val_accuracy: 0.5342\n",
      "Epoch 37/150\n",
      "6844/6844 [==============================] - 0s 14us/sample - loss: 1.1757 - accuracy: 0.5440 - val_loss: 1.2040 - val_accuracy: 0.5319\n",
      "Epoch 38/150\n",
      "6844/6844 [==============================] - 0s 14us/sample - loss: 1.1755 - accuracy: 0.5462 - val_loss: 1.2032 - val_accuracy: 0.5348\n",
      "Epoch 39/150\n",
      "6844/6844 [==============================] - 0s 13us/sample - loss: 1.1738 - accuracy: 0.5462 - val_loss: 1.2036 - val_accuracy: 0.5336\n",
      "Epoch 40/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1730 - accuracy: 0.5453 - val_loss: 1.2034 - val_accuracy: 0.5348\n",
      "Epoch 41/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1728 - accuracy: 0.5453 - val_loss: 1.2031 - val_accuracy: 0.5319\n",
      "Epoch 42/150\n",
      "6844/6844 [==============================] - 0s 13us/sample - loss: 1.1717 - accuracy: 0.5481 - val_loss: 1.2030 - val_accuracy: 0.5354\n",
      "Epoch 43/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1701 - accuracy: 0.5451 - val_loss: 1.2034 - val_accuracy: 0.5365\n",
      "Epoch 44/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1697 - accuracy: 0.5472 - val_loss: 1.2044 - val_accuracy: 0.5330\n",
      "Epoch 45/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1695 - accuracy: 0.5472 - val_loss: 1.2027 - val_accuracy: 0.5359\n",
      "Epoch 46/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1682 - accuracy: 0.5472 - val_loss: 1.2030 - val_accuracy: 0.5319\n",
      "Epoch 47/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1695 - accuracy: 0.5487 - val_loss: 1.2038 - val_accuracy: 0.5319\n",
      "Epoch 48/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1667 - accuracy: 0.5494 - val_loss: 1.2029 - val_accuracy: 0.5336\n",
      "Epoch 49/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1653 - accuracy: 0.5468 - val_loss: 1.2030 - val_accuracy: 0.5371\n",
      "Epoch 50/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1659 - accuracy: 0.5511 - val_loss: 1.2030 - val_accuracy: 0.5365\n",
      "Epoch 51/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1643 - accuracy: 0.5487 - val_loss: 1.2038 - val_accuracy: 0.5307\n",
      "Epoch 52/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1644 - accuracy: 0.5504 - val_loss: 1.2045 - val_accuracy: 0.5307\n",
      "Epoch 53/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1638 - accuracy: 0.5489 - val_loss: 1.2028 - val_accuracy: 0.5365\n",
      "Epoch 54/150\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1627 - accuracy: 0.5513 - val_loss: 1.2072 - val_accuracy: 0.5377\n",
      "Epoch 55/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1651 - accuracy: 0.5484 - val_loss: 1.2046 - val_accuracy: 0.5348\n",
      "Epoch 56/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1622 - accuracy: 0.5511 - val_loss: 1.2038 - val_accuracy: 0.5354\n",
      "Epoch 57/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1609 - accuracy: 0.5492 - val_loss: 1.2044 - val_accuracy: 0.5301\n",
      "Epoch 58/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1603 - accuracy: 0.5494 - val_loss: 1.2052 - val_accuracy: 0.5348\n",
      "Epoch 59/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1591 - accuracy: 0.5495 - val_loss: 1.2044 - val_accuracy: 0.5359\n",
      "Epoch 60/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1579 - accuracy: 0.5510 - val_loss: 1.2040 - val_accuracy: 0.5336\n",
      "Epoch 61/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1571 - accuracy: 0.5520 - val_loss: 1.2038 - val_accuracy: 0.5324\n",
      "Epoch 62/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1565 - accuracy: 0.5533 - val_loss: 1.2043 - val_accuracy: 0.5342\n",
      "Epoch 63/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1560 - accuracy: 0.5519 - val_loss: 1.2043 - val_accuracy: 0.5319\n",
      "Epoch 64/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1551 - accuracy: 0.5526 - val_loss: 1.2049 - val_accuracy: 0.5313\n",
      "Epoch 65/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1555 - accuracy: 0.5522 - val_loss: 1.2054 - val_accuracy: 0.5301\n",
      "Epoch 66/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1553 - accuracy: 0.5568 - val_loss: 1.2044 - val_accuracy: 0.5330\n",
      "Epoch 67/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1535 - accuracy: 0.5523 - val_loss: 1.2050 - val_accuracy: 0.5359\n",
      "Epoch 68/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1539 - accuracy: 0.5542 - val_loss: 1.2060 - val_accuracy: 0.5313\n",
      "Epoch 69/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1526 - accuracy: 0.5529 - val_loss: 1.2055 - val_accuracy: 0.5348\n",
      "Epoch 70/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1524 - accuracy: 0.5542 - val_loss: 1.2049 - val_accuracy: 0.5330\n",
      "Epoch 71/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1506 - accuracy: 0.5555 - val_loss: 1.2049 - val_accuracy: 0.5342\n",
      "Epoch 72/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1500 - accuracy: 0.5535 - val_loss: 1.2073 - val_accuracy: 0.5336\n",
      "Epoch 73/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1501 - accuracy: 0.5538 - val_loss: 1.2054 - val_accuracy: 0.5330\n",
      "Epoch 74/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1496 - accuracy: 0.5560 - val_loss: 1.2061 - val_accuracy: 0.5342\n",
      "Epoch 75/150\n",
      "6844/6844 [==============================] - ETA: 0s - loss: 1.1507 - accuracy: 0.55 - 0s 12us/sample - loss: 1.1482 - accuracy: 0.5555 - val_loss: 1.2053 - val_accuracy: 0.5324\n",
      "Epoch 76/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1475 - accuracy: 0.5573 - val_loss: 1.2062 - val_accuracy: 0.5359\n",
      "Epoch 77/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1480 - accuracy: 0.5574 - val_loss: 1.2069 - val_accuracy: 0.5307\n",
      "Epoch 78/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1479 - accuracy: 0.5541 - val_loss: 1.2066 - val_accuracy: 0.5342\n",
      "Epoch 79/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1461 - accuracy: 0.5552 - val_loss: 1.2074 - val_accuracy: 0.5389\n",
      "Epoch 80/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1456 - accuracy: 0.5567 - val_loss: 1.2045 - val_accuracy: 0.5371\n",
      "Epoch 81/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1463 - accuracy: 0.5563 - val_loss: 1.2073 - val_accuracy: 0.5324\n",
      "Epoch 82/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1453 - accuracy: 0.5554 - val_loss: 1.2098 - val_accuracy: 0.5336\n",
      "Epoch 83/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1446 - accuracy: 0.5544 - val_loss: 1.2061 - val_accuracy: 0.5365\n",
      "Epoch 84/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1441 - accuracy: 0.5570 - val_loss: 1.2064 - val_accuracy: 0.5395\n",
      "Epoch 85/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1437 - accuracy: 0.5580 - val_loss: 1.2071 - val_accuracy: 0.5400\n",
      "Epoch 86/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1428 - accuracy: 0.5573 - val_loss: 1.2078 - val_accuracy: 0.5354\n",
      "Epoch 87/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1414 - accuracy: 0.5601 - val_loss: 1.2073 - val_accuracy: 0.5336\n",
      "Epoch 88/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1413 - accuracy: 0.5603 - val_loss: 1.2071 - val_accuracy: 0.5377\n",
      "Epoch 89/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1413 - accuracy: 0.5593 - val_loss: 1.2073 - val_accuracy: 0.5359\n",
      "Epoch 90/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1400 - accuracy: 0.5598 - val_loss: 1.2072 - val_accuracy: 0.5371\n",
      "Epoch 91/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1400 - accuracy: 0.5595 - val_loss: 1.2090 - val_accuracy: 0.5400\n",
      "Epoch 92/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1394 - accuracy: 0.5590 - val_loss: 1.2077 - val_accuracy: 0.5406\n",
      "Epoch 93/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1393 - accuracy: 0.5603 - val_loss: 1.2074 - val_accuracy: 0.5383\n",
      "Epoch 94/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1381 - accuracy: 0.5621 - val_loss: 1.2076 - val_accuracy: 0.5359\n",
      "Epoch 95/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1390 - accuracy: 0.5606 - val_loss: 1.2086 - val_accuracy: 0.5395\n",
      "Epoch 96/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1370 - accuracy: 0.5601 - val_loss: 1.2070 - val_accuracy: 0.5412\n",
      "Epoch 97/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1371 - accuracy: 0.5580 - val_loss: 1.2087 - val_accuracy: 0.5348\n",
      "Epoch 98/150\n",
      "6844/6844 [==============================] - 0s 13us/sample - loss: 1.1366 - accuracy: 0.5596 - val_loss: 1.2071 - val_accuracy: 0.5395\n",
      "Epoch 99/150\n",
      "6844/6844 [==============================] - 0s 15us/sample - loss: 1.1365 - accuracy: 0.5637 - val_loss: 1.2077 - val_accuracy: 0.5330\n",
      "Epoch 100/150\n",
      "6844/6844 [==============================] - 0s 13us/sample - loss: 1.1372 - accuracy: 0.5596 - val_loss: 1.2117 - val_accuracy: 0.5283\n",
      "Epoch 101/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1356 - accuracy: 0.5582 - val_loss: 1.2075 - val_accuracy: 0.5389\n",
      "Epoch 102/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1346 - accuracy: 0.5631 - val_loss: 1.2088 - val_accuracy: 0.5348\n",
      "Epoch 103/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1345 - accuracy: 0.5611 - val_loss: 1.2076 - val_accuracy: 0.5348\n",
      "Epoch 104/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1337 - accuracy: 0.5624 - val_loss: 1.2081 - val_accuracy: 0.5301\n",
      "Epoch 105/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1341 - accuracy: 0.5620 - val_loss: 1.2082 - val_accuracy: 0.5371\n",
      "Epoch 106/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1341 - accuracy: 0.5617 - val_loss: 1.2094 - val_accuracy: 0.5348\n",
      "Epoch 107/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1334 - accuracy: 0.5631 - val_loss: 1.2084 - val_accuracy: 0.5377\n",
      "Epoch 108/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1329 - accuracy: 0.5656 - val_loss: 1.2084 - val_accuracy: 0.5365\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 109/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1333 - accuracy: 0.5631 - val_loss: 1.2090 - val_accuracy: 0.5336\n",
      "Epoch 110/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1328 - accuracy: 0.5633 - val_loss: 1.2081 - val_accuracy: 0.5377\n",
      "Epoch 111/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1311 - accuracy: 0.5639 - val_loss: 1.2093 - val_accuracy: 0.5354\n",
      "Epoch 112/150\n",
      "6844/6844 [==============================] - 0s 9us/sample - loss: 1.1301 - accuracy: 0.5644 - val_loss: 1.2089 - val_accuracy: 0.5359\n",
      "Epoch 113/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1299 - accuracy: 0.5624 - val_loss: 1.2095 - val_accuracy: 0.5389\n",
      "Epoch 114/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1296 - accuracy: 0.5640 - val_loss: 1.2102 - val_accuracy: 0.5342\n",
      "Epoch 115/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1293 - accuracy: 0.5646 - val_loss: 1.2095 - val_accuracy: 0.5348\n",
      "Epoch 116/150\n",
      "6844/6844 [==============================] - 0s 13us/sample - loss: 1.1289 - accuracy: 0.5630 - val_loss: 1.2101 - val_accuracy: 0.5348\n",
      "Epoch 117/150\n",
      "6844/6844 [==============================] - 0s 13us/sample - loss: 1.1297 - accuracy: 0.5640 - val_loss: 1.2096 - val_accuracy: 0.5342\n",
      "Epoch 118/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1288 - accuracy: 0.5615 - val_loss: 1.2094 - val_accuracy: 0.5342\n",
      "Epoch 119/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1284 - accuracy: 0.5653 - val_loss: 1.2102 - val_accuracy: 0.5348\n",
      "Epoch 120/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1275 - accuracy: 0.5639 - val_loss: 1.2099 - val_accuracy: 0.5354\n",
      "Epoch 121/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1271 - accuracy: 0.5637 - val_loss: 1.2111 - val_accuracy: 0.5330\n",
      "Epoch 122/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1264 - accuracy: 0.5650 - val_loss: 1.2095 - val_accuracy: 0.5301\n",
      "Epoch 123/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1263 - accuracy: 0.5630 - val_loss: 1.2106 - val_accuracy: 0.5342\n",
      "Epoch 124/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1270 - accuracy: 0.5650 - val_loss: 1.2103 - val_accuracy: 0.5330\n",
      "Epoch 125/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1259 - accuracy: 0.5652 - val_loss: 1.2109 - val_accuracy: 0.5348\n",
      "Epoch 126/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1262 - accuracy: 0.5668 - val_loss: 1.2119 - val_accuracy: 0.5336\n",
      "Epoch 127/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1259 - accuracy: 0.5647 - val_loss: 1.2110 - val_accuracy: 0.5365\n",
      "Epoch 128/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1260 - accuracy: 0.5652 - val_loss: 1.2126 - val_accuracy: 0.5319\n",
      "Epoch 129/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1252 - accuracy: 0.5647 - val_loss: 1.2115 - val_accuracy: 0.5342\n",
      "Epoch 130/150\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.1240 - accuracy: 0.5641 - val_loss: 1.2118 - val_accuracy: 0.5354\n",
      "Epoch 131/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1248 - accuracy: 0.5639 - val_loss: 1.2133 - val_accuracy: 0.5307\n",
      "Epoch 132/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1240 - accuracy: 0.5665 - val_loss: 1.2129 - val_accuracy: 0.5307\n",
      "Epoch 133/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1236 - accuracy: 0.5647 - val_loss: 1.2118 - val_accuracy: 0.5330\n",
      "Epoch 134/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1226 - accuracy: 0.5668 - val_loss: 1.2129 - val_accuracy: 0.5324\n",
      "Epoch 135/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1223 - accuracy: 0.5655 - val_loss: 1.2118 - val_accuracy: 0.5365\n",
      "Epoch 136/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1222 - accuracy: 0.5672 - val_loss: 1.2122 - val_accuracy: 0.5319\n",
      "Epoch 137/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1216 - accuracy: 0.5655 - val_loss: 1.2130 - val_accuracy: 0.5307\n",
      "Epoch 138/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1209 - accuracy: 0.5674 - val_loss: 1.2131 - val_accuracy: 0.5295\n",
      "Epoch 139/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1210 - accuracy: 0.5658 - val_loss: 1.2121 - val_accuracy: 0.5330\n",
      "Epoch 140/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1210 - accuracy: 0.5679 - val_loss: 1.2146 - val_accuracy: 0.5283\n",
      "Epoch 141/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1200 - accuracy: 0.5675 - val_loss: 1.2123 - val_accuracy: 0.5313\n",
      "Epoch 142/150\n",
      "6844/6844 [==============================] - 0s 12us/sample - loss: 1.1202 - accuracy: 0.5650 - val_loss: 1.2144 - val_accuracy: 0.5319\n",
      "Epoch 143/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1195 - accuracy: 0.5663 - val_loss: 1.2143 - val_accuracy: 0.5289\n",
      "Epoch 144/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1187 - accuracy: 0.5677 - val_loss: 1.2135 - val_accuracy: 0.5307\n",
      "Epoch 145/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1185 - accuracy: 0.5679 - val_loss: 1.2131 - val_accuracy: 0.5307\n",
      "Epoch 146/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1188 - accuracy: 0.5684 - val_loss: 1.2137 - val_accuracy: 0.5307\n",
      "Epoch 147/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1188 - accuracy: 0.5671 - val_loss: 1.2139 - val_accuracy: 0.5319\n",
      "Epoch 148/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1184 - accuracy: 0.5669 - val_loss: 1.2136 - val_accuracy: 0.5295\n",
      "Epoch 149/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1175 - accuracy: 0.5679 - val_loss: 1.2145 - val_accuracy: 0.5283\n",
      "Epoch 150/150\n",
      "6844/6844 [==============================] - 0s 11us/sample - loss: 1.1186 - accuracy: 0.5700 - val_loss: 1.2153 - val_accuracy: 0.5278\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(32, input_dim=46, activation='relu'))\n",
    "model.add(Dense(16, activation='relu'))\n",
    "model.add(Dense(7, activation='softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "history=model.fit(train_x, train_y_one, epochs=150, batch_size=500, verbose=1,validation_data=(test_x,test_y_one))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:19:39.723063Z",
     "start_time": "2020-03-01T09:19:39.154358Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABKAAAAHoCAYAAAB+apcaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nOzdd5xV1bn4/8/e+/TplWEKA9PovXdQBAGNLRg1isYkJjfRfGOiKWqKN/cmv5vkajQmGonmRsGKKCAovUnvMAMMDMMMTGN6OTOn7/374+DRcYbeBuZ5v178MefsstY++5T18KxnKYZhGAghhBBCCCGEEEIIcZmoV7sBQgghhBBCCCGEEOL6JgEoIYQQQgghhBBCCHFZSQBKCCGEEEIIIYQQQlxWEoASQgghhBBCCCGEEJeVBKCEEEIIIYQQQgghxGUlASghhBBCCCGEEEIIcVmZrnYDrpa6umZ03bgsx46LC6emxnlZjt1RdcY+g/S7s5F+dy7S72ubqirExIRd7WaIdlyu32DXy717vqTfnYv0u3ORfncu10O/z/b7q9MGoHTduGwBqM+P39l0xj6D9LuzkX53LtJvIS69y/kbrLPeu9LvzkX63blIvzuX673fMgVPCCGEEEIIIYQQQlxWEoASQgghhBBCCCGEEJeVBKCEEEIIIYQQQgghxGUlASghhBBCCCGEEEIIcVl12iLkZ+JyNeN01hMI+C9o/8pKFV3XL3GrOrZgnw0sFhsxMQkoinK1mySEEEIIIYQQQrRhGAZ1dVV4vW6gYxT+vlbiCJpmIjw8Grv9/FcblgDUV7hczTQ11REdnYDZbLmgQIrJpOL3d/wb51IymVR8Pj/19dU4nQ1ERERf7SYJIYQQQgghhBBtOJ0NKIpCly6pKErHmBh2LcQRDMPA5/NSX18FcN5BqI5xpTsQp7Oe6OgELBarZPGcJ0VRiYiIweVyXu2mCCGEEEIIIYQQ7XK5nERERHeY4NO1QlEULBYr0dEJOJ31572/XO2vCAT8mM2Wq92Ma5ammdD1wNVuhhBCCCGEEEII0S5dD6BpMiHsQpnNlgsqWSQBqHZI5tOFk2snhBBCCCGEEKKjk7HrhbvQaycBKCGEEEIIIYQQQghxWUkASgghhBBCCCGEEEJcVjLp8Trwpz/9nuXLPwEgEAjg8/mw2Wyh5998832SkpLO+XjLl3/CG2/8i7lz37vkbRVCCCGEEEIIIcS5u17G/BKAug48+eRTPPnkUwAsW7aUOXNeZv78xRd8vKlTpzN16vRL1TwhhBBCCCGEEEJcoOtlzC8BqE7g0UcfISmpK7t27UDTTMyd+x47dmzjzTf/RUnJCdxuF337DuDnP3+a5OQUli5dzOuvv8r8+YtZunQxixd/RM+evVix4lM0zcT48RN5/PGfYTLJ7SOEEEIIIYQQQlxN18qYXyII52Dj/nI+21d+ztsrChjGhZ9v3ICujO3f9cIP0I7du3cyZ86/sVisNDTU86tf/YL//u//YfTocTQ01PPUU0/y2mv/4Fe/+s82++7fv5dRo8awcOEyjh4t4NFHv8uwYSOYPHnKJW2jEEIIIYQQQghxpZ3vmP9SuNTj/mthzC8BqE5i5MjRxMXFA2Cz2XjzzXdJSUnF7XZTWVlJVFQ0VVWV7e5rtVqZPfthFEWhZ89eZGVlU1xcdAVbL4QQQgghhBBCiNO5Fsb8EoA6B2P7n19k0mRS8fv1y9ii8/f5jQhgMplYuXIZixZ9iGEYdO+egcvVjKa1fzvExMSiKErob00zEQgELnubhRBCXL/cXj8b9pYzqm8XIhyWq90ccY2rqndx4EQDfdKirnZThBBCXIPOd8zfEV0LY34JQHVCq1ev5P333+Ef//gXKSmpADz//B85erTgKrdMCCFEZ3Ci0skrC3OpqGmhR3KkBKDERdty4CQfbSjk1Scnoanq1W6OEEIIcVV11DG/fEN3Qk1NjWiahsViwTAMtmzZxKefLiEQ8F/tpgkhhLjCymuaKa9pJqCfe+aubhg0NHspqmhkx6FKFm88xquL8vjbh/s5UFSLcZpCiAFdZ83uUv7rjR20uP08cc8gslIkY6WjKCoqYsSIEZSUlLT7vK7rvPrqq9xwww0MGTKEhx56iMOHD1/hVrbPYTVhGNDskt8yQgghREcd80sGVCc0Y8at7N+/lwceuBtN00hP78HXv34PCxcuwO+XH25CCHGtqmvykH+8jmG9EjFpX/wfU3WDizW7SxndJ4nUxHAgGET6aMMxPt5UBICmKnSJdXDn5GyGZMa2ObZhGBwqrmPJlmLyj9cT0FsHmeIibfj8AXbmV5GRHMlNw9KIi7Jht2h4/TrbDp5kc95JGpu99O0Ry3dv6UNkmGQ+dRRr1qzh6aefpqGh4bTbvPHGG7z++uu88sor9OvXj7fffpvZs2ezdOlSYmPb3jNXUpg9+JPW6fLJfSWEEKLT66hjfsU43X9TXudqapzoetuuV1QUk5SUflHH7og1oC63L/f5UlzDa0VCQgRVVU1XuxlXnPS7c5F+n7sWt49FG4vYfaSKe27IZnBOwkW1Yf3eMiC4Sop6al6+y+PnzeX5nDjp5KbhaYzpl4SqKqzZVcoH647i9gZISQjjoem9yOgaycb9Fby18jBubwBFgRsGp3LTiDTeWnGYfUdrGNs/iV7dYiivaSH/eB1Hyxq5dUx3bh/fA0VR0A2DvQXVLNlcTGFZI1FhFkb3TSIuykZMhJW4SBtJsQ6sFg2fP8Bn+yv4ZEsx1Q3uVn3RVIUBmXGM7d+VQdnxof5cLqqqEBcXflnPcb14+eWX+fjjj3nooYd45plnWLVqFampqW22u/vuu5k4cSI//OEPQ49Nnz6dBx54gPvuu++cz3e632AXI/dYDc+9u5dffHMIOWnRl/TYHZ18Rncu0u/ORfp9+XTEMeu1Fkdo7xqe7feXZEAJIYS4otxeP0s2FzOsZyLpSRFXtS0nKp3ERVpx2MwXtL9uGHi8AdzeAPsLa/hg3VGcLT7iomz8dcF+pgxLZdakLMwmlYCu0+D0EhlmaZWd1B7DMFiwvpAlm4sB+GxfOQ9O74XPH+CVj/KobnCTFOfg/z45xMLPjhEVZqGooom+3WMY1TeJBesL+f0bO+mWFEFxRRM5adHce2M26/eVsXp3Cat2laCpCg9MzWHS4JRQ0cmArvPe2kIWbyqiodlDr/QYlm4upqSqmfgoGw9M68m4/kmYTVq77TabNCYPTmHCwK4UlTfR4vHj8vjRdYO+PWKl1lMHdeedd/L973+f0tLSM26n6zp2u73VY6qqUlhYeF7nuxyBwTRPsFCqZjGRkHB1P1euhs7YZ5B+dzbS787lcve7slLFZOp4FYk6YptOR1XV836dJAAlhBDiimls9vKX9/dSVNHEyh0l/PDOfvTrEXfGfby+AMcrnUQ6zCTGONo8bxgGdU0eCssacbp9jO6ThNXSfoDkyw4W1fLnd/bgsJmYMSqdG4amYjV/sV9AN9h64CSfbCnGatH47i19iI8ODr4/z0LaeuAkX84jzkqN4id355AcH8b7awpYuaOE3MJaNE3hZG0L/oCBqigkxNhJjnOQ3iWCHsmR9OgaSbjdHOrP2yuPsHJnCRMHJZOVEsU7q47w29e3oSgQ4bDw828OJislirxjtXy8qYjqRjffuaU3o/smoSgKQ3ISWLCukI255dw9OYupI9JQFYUHknoyaVAKK7afYPzArmSnts4U0VSVx+4ehNWk8PGmYtbvLSc5Pozv3tKHEX0Sz7m4s6aqZEptp2tGly5dzmm7adOm8eabbzJmzBgyMzOZP38+x44dY8iQIed1vsuRAeVzewEorWikqkvnynyTDInORfrduUi/Lx9d1ztcttG1lgGl63qb10kyoIQQQpzVniPVJMc72g3wXCqV9S6ee3cP9U0evjWjFyu2l/DC+/t4eGYwcALBjKLymhaOlTVyrLyRwvJGSiqdBHQDTVW4+4YspgxNRVEUPN4AS7YUs2FvGQ3N3tB5Fm8s4u7JWYzonUhVg5tN+8s5WFzHnRMy6NktBoCmFi+vfnyALrEOEmPsvL/2KMt3nKBPeix2q4bFrLH3aA3l1c0kxTqoanDx239t5+GZvYmLtPHywlyq6l1MHpxCfJQdm1UjPtJG3x5fLGF730059EqPYcnmYqLCLAzIiCM+ykad00t5TTNl1c3sOVLN58PwcLuZ6HArJk2hqKKJm4alcc+NWSiKQv/MOOavOYrXH+D+qT1Dwap+GXH0y2gbwLNbTXxzag733ZTdakldgLTEcB6e2fu0r5OiKNw5IZNuiRGoqnJFpsuJa8PDDz+M2+3mBz/4AV6vl+nTpzNu3Diioq5+sPHz90Szy3eVWyKEEEKI05EAlBBCdHIHimp58YN9dImx89uHR7TKAvqyVTtLOFJSz71Tcoj6SpFfwzDaBDo+p+sGm/MqeH9NAQHd4Il7g9k7Q3MSeWnBPuYsPsD8tUdxe/24vYFQRpHNotGjayQ3j+xGepcINuVW8PbKIxwqrgtm+KwvpK7Jw+DsePp0jyUjORKvL8Dbq47wj0V5zF97lJpGNwoQZjfzv+/u5ZFb+zC0ZwKvLzlIs8vH47MG0q1LBPnH61i8qYjDJ+pD7eiREsUP7+jH4JwEqutdvLwwj5cW7EdTFSLDLPz8vrPXmhmSk8CQM9SBcnn8FFU0UVTeSFWDm/omD/VOD3dNzGDGqPTQNY10WM4YNDqd070m52JYr8QL3ldcnyoqKrjrrrt47LHHAPD7/dx4443cddddV7llYDVrmDQVpwSghBBCiA5LAlBCCHEd2FNQTV5hLeMHdqVbl3Ofi+3y+PnX0oNEhVs4Wefig3VHuW9KTpvtahrcvLu6AH9AJ/94Pd+9tQ+902PIP17Pks1F5J+oJz7KTtc4B0lxDmIjggWqfX6dxZuKKKtuJj0pgu/e0ofk+DAAHDYTj989kEUbi2hwerFZNGxWjcRoBxnJkSTFOVpl3gztmcCKHSW8v6aA3UeqSU+K4D9u60dWauvsi18/OJwN+8rYdrCSSYOTGd03CYtZ44X5e3n5o1wGZcez92gN907JDl2rnt1iQtlRn/ty+ndijIOn7h/KRxuCQa97p2RfknpGdquJ3ukx9E6POfvGQlxlS5YsYenSpfzrX//CYrHw0ksvYbFYmDx58tVuGoqiEBlmlgCUEEII0YFJAEoIIa5hDU4P81YeYcehShRg1a4SRvbpwu3je9DlHKbTvbu6gNomD7+8fyhb806yckcJQ3MS2gRjFqw/iqLAj2cN5N3VR/jfd/aQkhBGSVUzkWEWJg5Moc7pobymmX1Hawh8qb5LUqyDH9zej6E9E9pk5JhNGndNzDynviqKwtThafRMi+ZkXQvDeiaiqm0zfFRVYeKgFCYOSmn1+BP3DOaVj3LZfaSaAZlxTBnadpWvMzGbVGZNzjqvfYS4lpWVlTFz5kzmzJnDsGHD+Na3vkV5eTnTp0/H7/czfPhw/u///g+LpWMUl49wWCQAJYQQQnRgEoASQogryDAM9hyp5kSV89R0Ky9pieGM7Z/Uqv6SPxAsQHi61dLcXj9rd5fx8aYivP4Ad4zvwcRBKazYcYIV20+w9cBJkmId9OgaSVZKJGP6d20ztS63sIb1e8uYPrIbWSlRpCWEs7+whteXHuTZh0dgswS/IgpO1LM57yQzR6czIDOOnmnRvLXyMIdLGrh/ag7j+nfF8qVj64ZBU7OXOqcHtydAdlrUORevPhfpSREXtHqe1azx6F392X6wkgGZcRc1PU2I61Fqair5+fmhv5OTk9m9e3fob7PZzG9+8xt+85vfXI3mnVVEmEVqQAkhhBAdmASghBDiCjEMgwXrC1myuRgIFs2NcJjZW1DN4k1FZKdGEW43U17TQlW9i4BuEOkwEx1hDU1v6xrnoKrezcodJ2h2++nbPYb7bsqha1xwWttdEzO5cWgqG/eXc7S0kbyiWjbnVbB6dynfv60fKfFhoSDYv5fl0zXOwe3jewBgtWg8PLM3/zNvF3/9YD/3TskmJT6M1xfnEeEwM2NUemi7b804fT0iVVGICrcSFW69zFf0/GmqyqhTBc+FENeXCIeF4gb31W6GEEIIIU5DAlBCCHGZeHyBUNaRYRh8uCEYfJo4KJn7pmRjNgWfq210szmvgq0HTuJ0+UiJD2NozwRMmkpdk4e6Jk9oxTT9VIXuQVnxzByd3u5S99HhVmaO7h46b96xWuZ8fIDf/Xs7d4zPIPdYLXnHakmOD+P7t/UNtQMgJy2a+6f1ZP7aAn7z2jb6dI8hr6iO+6fmYLfKV4YQouOKDLPgbPGefUMhhBBCXBUymhBn5HQ6CQT8REWdeaUnITo6wzAoLG+krtGDy+vHHzAYmpNAZNilrV3S7Pax7cBJNuZWUFjWSGKMnT7dYwFYu7uUCQOTeWBaz1bFtWMjbcwc3T0UNDodf0Cnss6FpinnVN8JgnWT+mXE8ezDI3h1UR7vri7AbjVx75RsJg9OaXeK3+TBKQzvlcjSLcWs2llCamI4EwYmn/tFEEKIqyAyzEKz23/GVTmFEEII0dqVHPNLAOo68Kc//Z7lyz8BIBAI4PP5sNlsoefffPN9kpLOf8qJruvcc88d/Pd//4mBAwddsvaKzsvjCzB3eT4TB6WQ1U7mzuVS2+hm7vLD7CmobvX42t2lPHX/UKwWrc0+Pn+A9XvLqW5w0TUujK5xDsIibG22+7INe8t4c/lh/AGdlIQwZo5Op6TSyea8CjzeAOMHdGX2za2DT+fDpKmhFeTOV3S4lSfuGczuI9Vkp0UReZYV3MLtZu6enMXNI7uREB+Bp8VzQecVQogrJcJhIaAbuDwBHDb5iSuEEOL6cb2M+eXb+Trw5JNP8eSTTwGwbNlS5sx5mfnzF1/0cXVdp76+7qKPI8TnPtlSzMb9FRwpaeB33x7RaurXuah3emh2+UhJCD/tNnnHavlkazGRYRa6xoWhAEu2FGPoBrMmZ9KvRxw2i8aJSid/+3A/ry05wH/c3i/0v+UBXWfT/goWbjxGbaMHk6bgDwSnvWmqQq/0GIbmJDA4J4GoL2VPrdhxgrdXHqFP9xhmTcqiW5fw0DH9AZ3aJg8JUbar+r/yqqowtGfCee0T6bAQGWahSgJQQogOLuJUYN3p9kkASgghxHXlehnzy7fzOfAd3ogvf/05b68oCoZhnH3D0zD3nIA5Z+wF7/9VXq+X119/lZUrl+FytTBgwGB+8pOfkZCQCMBrr/2Djz9eiNvtJjU1je985/uMHDma++67C4Cf/OSH/PCHP+bOO2ddsjaJzqe63sUnW4/TLTGc45VOPtlynK+N63FO+1bWu/h0SzGf7S8noBvcNq4Ht4zujqp+EcwxDINPtx5n/rqjxERYqahtYUveSQD69ohl9rSeJETbQ9snRNuZNSmL99YUsHhTEVOGpvLZvnJW7yqlst5Fj66RfHtGb3p2i6Gm0U15TTPHq1r4bG8pbyzL583l+fTrEcfY/kmcrHPx4fpChuQk8L2v9cVsaj2tzaSpJH7p3EIIIS69z6dUN7t8IJ+5QgghzsP5jvkvhUs57r9WxvwSgOoE/v73F9m/fy8vvfQqMTExzJnzCk8++WNee+1Ndu3azkcffcDrr88lPj6BDz+cz3PP/Q/vvPMhb731AZMmjeK55/4mU/DERXtndQGKAj/6+gDeW1PAx5uLGdW3C4kxDmob3Xyw7igWs8bNI7uF6huVVjlZuqWYrQcqUVUY178rbl+AjzYcI/94PQ9N74WuG9Q1eVi9u5QdhyoZ3iuRb83ohc1iwu3109DsJTHa3m7m0bQRaZyodPLRhmMs3VKM16eTlRrF3TdkMTg7PrRPQrSdhGg7N46KYObINEqrmtl68CSbcit4ZWEeAKP6duHbM3ujqW1rKgkhhLj8QhlQLt9VbokQQghxZV0rY34JQJ0Dc87Y84pMmkwqfr9+GVt07nRdZ9GiD/njH58nKakrAN///qNMnz6Zffv2YDZbcDqbWLJkEePGTeT22+/ijju+LsU7xVnVOz0s336CEyebuGFoKoOy4lvdN18uApt3rJZdh6u4c0IGsZE2vnFDNnuP1jBvxRFG9+3C3OWH8es6ug7r95YxvFciPr/O7iPVWM0aNw1PZerwbsREWDEMg97pMcxbfpifv7I5dD5FgVmTM7l5RLfQeW0WEzbL6T/mFEXhoek98fkD2CwmbhyaSnpSxBn7rSgKqYnhpCaGc8f4DA4er6OqzsWEgcmtMrKEEEJcWRFhZkACUEIIIc7f+Y75O5Jracx/RQNQ1dXVPPPMM2zfvh2bzca9997Lo48+2u62t99+O4WFhWjaFzVi5s+fT2ZmJocOHeL222/Hbv8ivbpPnz7MmzfvsvfhWlNbW4PX6+Gpp55sNTjWdZ3y8jJmzLiVZ5/9Ax988B5vvPEvwsPDueuuu5k9++Gr2GrRkZXXNPP++kJWbD1OQNeJDrfy1w/2k5kSyY1DUymrbuZAUR3FFU3ERFjpGhdGWXUzidF2po3oBkBMhJU7xvXgndUF7C+sITMlku/c0gebWWP59hOs3l2Kpih8bWx3pgxLI9xuDp1fURTGD0gmKyWKvQU1RDjMxERYSYp1EBt55iLh7TGbNH5wR/8LuhaqqtC3eyx0v6DdhRBCXEKSASWEEKIzupbG/Fc0APX444+TmprKxo0bOXHiBI888gjdu3fnlltuabWd1+uloKCARYsWkZGR0eY4+/fvZ+DAgbz77rtXqunXrOjoGEwmEy+++Aq9evUOPV5YWEBKSioVFRUkJnbhhRf+jsfjYevWzfzmN7+kV68+DB06/Cq2XHxVY7OXmkY3PbpGntP2LW4/n+0rY0z/rq0COAD5x+voGhcWqpdxLsfaeqCCjbkVFJY1YtIUxvTryvRR3YiLtLFxfzkLPzvGq4sOoCoKGcnBYFS900NFTQtur79NbaQbh6VSfLKJlIRwpo1IC01dmzU5i6+N64GqcMYi5cGV6S5sRTghhBDXn3CHBQVwtkgASgghROdxLY35r1gAqri4mG3btvH8889js9nIzs7mwQcfZN68eW0CUAcPHsRqtdKjR/sFinNzc+nXr9+VaPY1z2QyMW3aDP7xj5d4+ulniYmJYcGC93n55ReZO/d9Dh06yHPP/Q8vvPAymZlZJCQkYBgGUVHRmEwmNE2judl5tbvR6em6wV/e30txRRP3TMnmpmFpoedKKp0cKK5jZO9EosKtAFTVu3hh/j7KqpvZfOAkP7t3MHZr8O3+6dbjvLemgHC7mQdv7snQnomnPa/XF2DVrhKWbi6m2e0nJSGMuydnMXNCJgHPFz/wJw5KYXTfJIoqmkhLDA+d60w0VeW7t/Zt9zmr+fxWxxNCCCE0VcFhM+F0SwBKCCFE53EtjfmvWADqyJEjREdHEx8fH3osIyOD/Pz8Ntvm5ubicDiYPXs2hw8fJiUlhUcffZQbbrgh9LzdbmfatGk0NTUxfPhwfvnLX5KUlHSlunNN+X//7wn++c+X+d73HqKpqYn09O786U8vkJycQnJyCsePF/Gzn/2YhoZ64uLi+elPfxGKnN5yy2386lc/Z/bsh3nwwW9f5Z5c/1rcfo6U1NMl1kFSrCP0+KqdJaHgztsrj9Dg9PK1sd1ZvKmIT7ceJ6AbLFh/lKnD08hJjWbOxwfQdYM7xvdg0cYiXpi/j8fvHsi63aW8t6aAwdnx1DZ5+NuHuYzum8Sovl1wewO4PH7c3gBuj58Wj5/thyqpa/LQLyOW28dl0KNrBIqiEBtpo6qq9Q98i1kjJy36Sl8yIYQQAl/RTspXrifcPiq4Cp4QQgjRiVwrY37FMAzjsp7hlIULF/L888+zdu3a0GPbt2/nwQcf5MCBA622nTt3Lps2beKJJ54gNTWVFStW8Itf/II333yTQYMG8cgjjzBw4EAeeOABAoEA//Vf/8WRI0f48MMPW9WMuhB5eQdITk6/qGN0dmVlxfTt2+dqN+Oa4fL4WbrxGJtzyzlyoh5dN7BaNJ751ggG5SRSVefih39aRe8ecfz626N4ZcE+Pt1cRJjdTLPLxw3D0pg5tgcL1x1l/Z5SALrGh/Gb74wiJSGc9btL+PO8naQmhnPipJOxA5N58ptDMYB3VuTz/qoj6HrbjwGrRSMrNZpv3tyL/pnxbZ4XQgghLkRNjbPd752L4dnzMd5t8/m79XtoVhtP3DP4kh6/I0tIiKCqqulqN+OKk353LtLvzuVK9LuiopikpI417u9Ii5mdi/auoaoqxMWFn3afK5YB5XA4cLvdrR5zuVyEh7dt3P3338/9998f+nvmzJksWrSIZcuWMWjQIF599dVW2z/99NOMHj2ao0ePkpOTc07tOd2PH13XL/pFv9ZunEvhy33Wdb3TfFB+/uG49cBJlmwu5tsze591FbXP+QM66/aUsXjjMRpbfGQkRzJjVDpZKVHMX1vAs//cyg9u78eGfWUEAgbfmJRJbY2TWRN64DCr7Miv5Ptf60vfHrEAPHRzTyYPSmbX4SpuGp6GBYOqqiZ6p0Yxe1pP/v1pPkN7JvDg1Bxqa5sBmDY0lcEZsTS2+LBZNOwWE3arhtWihWoyAW1eT/ky7Fyk353L9dLvs/0AEtcXxRb87o21+SmVGlBCCCFEh3TFAlDZ2dnU1dVRW1tLbGxwwHz06FGysrLabDtv3jxSU1OZOHFi6DGPx4PVaqWmpoZ//vOffO973yM6Ojjdx+v1AmCznf8KWEJcrG0HT/Lq4jww4H/f3cMv7x8SKo69M7+St1cdweXxt9nPHzDw+XV6pkXz2F2ZZKZEhZ7LSI7k+ff28NcF+zAMmDU5k4To4KqPiqJwy5ju3DKme5tjpidFtBsAmzgohd7dY4mPtLVaGQEgMcZBYszFXAEhhBDi6lKswWBjjMXP4Rr1LFsLIYQQ4mq4YgGo7t27M3jwYP74xz/y61//mrKyMt544w1+8IMftNm2srKS119/nddee43U1FQWLlzI7t27efbZZ4mKimLZsmU0Nzfz1FNP4XK5ePbZZxk9ejTdunW7Ut0R1zF/QGfBukLyT9SFHosKszIoO55BWfGtVo7bsDWXEXkAACAASURBVKeUVxcdIDs1mvumZPPce3v58zt7eOKeQazeWcqqXSWkd4lgaE7bQt+KAv16xNK3RyyK0jooFG4388Q9g3lpwX68/kCrouMXKvFUAEsIIYS43ii2YAAqyuTD6ZIAlBBCCNERXbEAFMCLL77If/7nfzJ58mTMZjP33Xcfs2bNoqysjJkzZzJnzhyGDRvGo48+it/vZ/bs2TQ2NpKZmcmrr75KenpwfuGcOXP4wx/+wIQJEzAMg0mTJvGHP/zhSnZFXCd2Ha7CbFLp2z0WVVWod3r4+4e5FJQ20KtbNGZTsKZYSZWTPQXVKAokx4ehqQoYUFLdTGZKJD+eNQCbxcRPvzGI/5m3i2f+uRXDgKnD0/j6pExM2vn/GLZbTTx572B0w0D9SoBKCCGEEF/4PAMq0uTF4zPh8wdC3+FCCCGE6BiuaAAqMTGRl156qc3jycnJ7N69O/S32WzmySef5Mknn2z3OJmZmfzzn/+8TK1UMAwdRZH/PbsQl7KmfV2Th8ZmL926hLfJEDpXjc1eVuw4wWf7ypk8JIVbx3QPHevTrcd5b00BAFHhFob3SmT7oUpcHj/fv60vI3p3CR3HMAxOVDrZmV/FicovlqjsmxnPLaO6YbME30ppieE8/o2BvLu6gBkj0xmUffHFuyX4JIQQQpyZYgtOfQ9XPYADp8tPTIQEoIQQQpyeYRgXPM7s7AxDB87/2l3RANS1wGKxUV9fTUREDJpmkhvyPBiGQXNzIyaT5ewbn8WWAxX8+9N8PN4AXWIdjO2XxKCseBw2E3arCZOm4PIGcHsDGIZBQrQ9FKgxDINj5U1szqtgw94yfH6d1MRwPtpwjJO1LTw0vTdrdpXw3poCRvROZHivRDbur2DNrlJiI6389O5hpCa2LlyrKArdukTQrUvr+krtFevNTI7iqfuHXvQ1EEIIIcS5+TwDyoEHAKfLR0yE9Wo2SQghRAdmMllobm4kLCxSxvznwTAMAgE/TU11WCznX4NbAlBfEROTgNPZQG3tSXQ9cEHHUFUVXe9cq+B93meTyUJMTMIZtw3oOh+uP0ZqYhgje3dp9Yb3+AK8vfII6/eWkZUSxeh+SWw9cJIF6wtZsL7wtMe0WTS6J0UQH2Unr6iWuiYPmqowqm8XZoxKJynWwcebi/lwfSGFZY2crHMxrFci3721D5qqMrRnIi6PH7NJvaDpckIIIYS4ehTNhGJ1YDVcQDAAJYQQQpxOTEwCdXVVOJ31V7spIddKHEFVNez2cMLDo86+8VdIAOorFEUhIiKaiIjoCz7G9bKE9fk41z4bhsEbn+azYV85ALvyq3hgWk9sFo3P9lfwyZZiqhvczBydzm3jemDSVCYPTqGy3kVReSMujx+XJ4A/oGO3mrBZNHTdoOhkE8fKGtl9pIqctGjunJDBwKx4wu3m0LlvHdOdLjF2XltykGE9E3jkVPDpc3arvB3E9ccwDIyWetQwWepQCHF90+zhWHU3AM0SgBJCCHEGmmYiPr7r1W5GK50hjiAjbnFBTta2sHZPKc1uP26PH7PFRFZyJIOz44kKs1BY1siqXSXsOVLN4Ox4vj4pi5gIKx9tOMaGfeXMHJ2OzaLx0YZjHC5pQFGgweklIzmSb03vRe/usa3OlxhtP+MqbuPPsd0jenehf0YcNosmqZbimmL4vei1JwhUFqJY7Jiyx571HtadtXg2vom/eDe2qY9h7i5TQ4UQ1y/NHoEp0AJIBpQQQgjREUkASpy3o6UNvDB/H26vnwiHBZtFI2DA5v3lzF2WT2ykjZpGN3arRt8esWw/VMmuw8FA1JYDJxk/oCt3TshAURT6Z8Qxb8VhLCaVR27pQ6/0mPMKDBm6juGsQY0887S/L5NMJ3G5GIZOoPQggcoCApWFGE012Kb8AC0m+aKO6dk4F9+hdfClacGmkjxsE76F0k7NNcPQ8R1YjWfbfNB1FGs4vn3L2gSg9KZqFEc0iibvCSHEtU+1R6A1NgASgBJCCCE6Ihl1CJrdPhZuOEZMpJWbR3QLBYAMw2DtnjKOljbQPyOOAZlxHD5Rz8sf5RIdbuWZ2UNJjHEAEB8fzp6DFew6XEVhWSMzRnVjVN8k7FYTlXUtvLu6gC0HTjIwM47ZN/cMnaNblwh+eREFu3356/Fs+D9sU3+EufuQi78YotPzFW4DFEw9hp1fMNTQca97Hf/hzwAFNaYrurMOz8Y3sc/82QVl3BmGgWfjPHwHVmPKGY8pfRBaQg98hz/Du2MBrqZqbFMfQ7VHttrPu/MjvLsWoaX0xTb+QfxFO/FseZdAdTFafDoAgboyWj74FZYB07GO+Ppp2+AvyaW5zgQxvc67/ZdL4GQBemMl5uwxV7spQogORLNHQFUpFrMqASghhBCiA5IAVCfj8wczKMwmDcMw2H6okrdWHqGx2QvAiUon35reC0VRmLv8MOv3lmE1a2zKrcCkKeg6pHUJ5/FZA4kM+yLzQlEUUhPCSU0Ib3POxBgHj901gNIqJ11iHa3qLl0sf/EeANxr5qDd8RvU6KRLduxrnaHr6LUnUOPSUBQprH4uArWluFe9AoaOljYA27jZqBHxZ93PMAw8m97Cf/gzLINvxTJwBorFjjdvJZ6Nc/Ef24E5Y3i7++qNVSj2CBRz21UkvNs/wHdgFeYBN2Md+Y1QEMs65GuoUUm4186h5aPfYZ/+OFp0MMvKX7wb765FmHLGYZv4bRRFwdxzAp4dH+LLW4k28dun2jsP9ADeg2uwDLkVxdT+alHuz97E1VyD485n0WJSzvVStitQcwI1OglFM592G91Zi2Kxo1jan3LrL8/HtfR/IeBFjUxE65J1UW0SQlw/VHsEhsdJhN0sASghhBCiA5IAVCdyrLyR597dQ7PbT7jdzChbASedCjGxfXh81kByj9XwwbpCquvdmE0qB4vrmDk6ndvH96CwrJGd+VUEdIM7J2Rc0DS2lHaCUxfDCPgJlB/ClD6YQMURXCv+iuP2X7U7kL9W+Yv3EDh5BDWhB1pCBkrY2acoGm4n3kPr8B1YjeGswdznBqxjH5CaV2dhGAaezfPAYscycDreXYtpfv8pbOMexJwz9oz7encswJe3EnP/aViG3Rm61ubek/EdXIdnyzuYug1oE+QJ1JbQsuC3KPYIbGNnY+o+GADd1Yh371J8+z7F3GtSq+DT58yZI1Aj4nAteyEYhLrpMdTwOFxrXkWNT8c2bnZoH8Uahjl7LL7DG7CMvJtAeT6B0jxMmSPxH92Kr2ALll4T2/RLd9ZgNJ4EwL32nzhuewZF1c54LfSWBnwH12LpNwXFGhZ63LPnY7zb5qPYIjD3noS592TU8Na13nRnDc3zn0GxOLDf/DhabGrr61VdhOvTv6CEx4LPjXvTvOB7XgKsQghOZUB5XUTYJANKCCGE6Ii03/72t7+92o24GlwuL4ZxeY4dFmalpcV7eQ5+gUqqnPz57d3YrSamjUgj2eFjRtP7DLUWMS7VR1xWf3pmdiUlPozVu0upbnDz8IzeTBvRDVVRiIu00e/UNDyzqe1g73z6bHhdtCz+AxgGWkL3027n3vIO3r1L0VL7thtUClTk4z+0DsvIuzFnjcKXuwy9ofK8p05djK/2O1B7Au+OD3Fvmosan94meyZQV4pidZzTgDlQdQzXJ/9LoDwff+E2fPuXESjPx5Q95rT9C1QW0jz/VwRK9qPFd0fr2hP/ofWg+zGl9D2vvumNlRheF4rVcdZ+Xyi9vgJMlrMGNS6VQH0ZijWs3evnP7YD396lWEd9A+uAmzFnjUavPIrv4BrMWaNRrGHt9tu79xO8OxZg7jWhTaBPUVTU6GR8uStA0TAl9w49Z+h+XJ8+DwE/ij0CX+4K9NoS/Mf34ln3OnrFYcw9x2Md/xDKabIG1bBYTBnDCRzfh2//cvxFO0EP4LjlZ22m5SmRCfhyV6JoFrx7l6KGx2Gf8QT+ot3olYWYe09uc138x3biL95N9LhZuA5tBNWEqWvPM15jz/YF+PYuwV+8G1PaABRrGN68VXi3vIMpfTBqeCy+Q+vx5a5AsYWjJWYEr4dh4F71d4ymahRFw3dwDVp8OmpUFwzdT6DiCO5lL6BYbDhu/QVqdFd8eStRwmPR4rufsU0X6lLc5/7je3At/ytaUg6q4/yXqj0X3gOr8Wx6C1PmiEtSz6sjfoddCEVRcDja1kgTV9/l+g1majmJ6+gujjgG0+RVmTDwwuvvXUuul/fs+ZJ+dy7S785F+n3tOtvvLwlAXQYd7caprGvhj2/tRtUUfv7NIQzJSSDbtYdAaR7mAdPxH92G7+Bq1KgupGXnMDQngQkDk+mXEXfO5zifPns2vUWgeDd6XSnmflPaDQYEThbgWf86RlMV/qPb0FL6tBm8+Q6uJXCyENv4B4NFnlUtOMUouRdqRPtFyQ2/F++ODzGaqlGju1504OPzfuuNVbiW/xXvtvfR60pRFBV/4XZMWaNCU4m8B9fi/uQ5dGftWetV6e4mXEv+iGK2ETbr95gzR4LFgf/oFtTY1NNOhXKvfx3cThy3/xrrwBmY0odguBrx5S4HRUWx2PEX78aXvwElLKbNNTUMncCJvbg3zsWzaR6+/PVocd1Qo7q02++L4Tu2E9fH/x96fRnmzBEXdaxzOt+RTbiW/BEUBVNy63pGht+Da9kLqOHxwcLeiopidaCl9sOXtxqj8STmzJFt+u09uBbPpnmYMkZgm/iddgNFakQ8en05vvx1mFL6hLJ+vLsX4y/Ygm3Sd7GOvhc0M75D69AbTmLuOR7b5O9i6TXxrMFUxerAnD2aQE0xek0x9pseQ0vo0bYd9kgCFYeDNaq8Lmw3/gdaZCKoGv5D69CSe7cJmHr3fYrhaqTrPU/jLC3Cd3ANpvQhpw2kGH4v7rVzUGNT0Ztr8edvwPC68G57H1P6YGw3/RBz9hjM2WPRG0/iy12OGhGPFtcN36F1+HKXYx19H9ZR3yBQsj8YVDu+F8/mt/AfWodiPhV8ikxEjUklUHYQ/9FtmHtNbLcY+7kwDB29vozA8b34Dn8GiooamQhc/H3uLzuIa9mLGC316HWlmHLGtf95V1uCd+8nKNZw1LDo8zpHoLYU9/K/YjirATCl9Lng9n6uo32HXSgJQHVcl+s3mMVbR/OhLRyPGEBJo8KUYWmX/iQd0PXynj1f0u/ORfrduUi/r11n+/0lU/Cucx5vgD+/s4eAbvDz+waTGB0MhviObEZNzMA26htY+kzGtfJveDa+ianbQJLjw85y1AvnL8nFd3ANalwaes0JAmWH2gyYDF3HvXEuiiMa243/gXvVy7Qs+m/sN/4AU7cBXxzrRC5aUlYowGPpPxXv7o/xH93aKtPkc3pLPa5lL6JXFQKgbH0Pc68JmPtPazWgNgI+XCv+htFUhbnPZMzZY8FkCQZu8lZhuBqwT/8pangwQGf4PbiWv4jurMYyYhaWXhPRW+pp+eg/ca38G45bfhEsAr3h3yj2SPyHP8PfYxim9EHtXiND13GvegXD1YDja0+jhsVAWAxqfHf8RTvx7l3abpZXoLqYwIn9WIbfFZq6pCgK1nEPYPjceHcswLtjwZe2P4bjtl99qei8jmvJnwiUHURxRGMZchv+4j24lj2PdfQ3sfSb0ratXhd6fXkoi+Vc+Utyca96Gcw2/Md24C89ELoPDN2Pa9mL4PdgHTsbLfaLYJvurEFvrESL69ZqelerNvnc6LUlqAkZoYBQoOY47vX/B4qKd9+nmPveiGqLCO3j3bMUw1mD7dZHWgUl1bAYLINvxbt9Pv6SXEgYHXrOV7AFz4Z/B2tFTX7ktFlKANaR3yBQcZiWRf+Nue9NmDKG4925CFPmqFBtKOvgW7H0uSEUKDwfisWO/eYfY7gaUR2nD2BY+t6Eq/QApowRofeIOWsUnq3v4ctb2SowZxgGgdIDaMm9Q/dRoPwQLR/8Gk5l8Knx6Thu/UUo+OMv2gWeZqzD70INj6dl2fN4dy9GS+6N7cb/QFGDXzlqZAL2mx7F9elfcK97DcPTgmfHArSUPpj7TEZRVBy3PoV749xT78Mb0BIzgoHoU6+boihYx3yTlg9/i2vNHFRHJIHKwmBW3ecjW01DjeuGlpARLL5uPjUFMuBHrz1BoLKQQNUx8LlPXUgVX+5KrKPvwdxvarvXUG+px2hpQI370oINbifuLe8SKD+IOWME5j43YLgag0HNyARM2WPwbpuP/+gWzFlf3EP+4t149y0jUH4o2Kzi3Ti+/l/nnMVk6H7ca+egWOxoXbKC93bP8W0CxkJ0Fqo9+PkQbfbjdMnUXCGEEKKjkQyoy6AjRS63HjjJxv0VPHbXADJTgkGWQM0JvLs+wjJoJlpiBoo1DDWqK74Dq1Ac0acNJviP78V7cC1afPc22Qbn0mfD24Jr6f+iOqJwfO1pvIfWgdfVpjizL389/oNrsY1/EHP6IEyZIwmcyA0WY84eg2J1oLc04N36Dubek0JTghTVhF5bQuD4Psz9p7Wa5haoOYFryR8xnNXYpvwQc+9JwcygwxvwH90SzP5wRGPoAdyr/k6geDeKPQr/4c/w5q3Cd3At/kPrwDAwXE34i3ZiyhxBWGQEtUtfJlB6APvUH2HpOQ7FZEG1R6JGJuLbv5xAVSG+fcvQkrJw3PY0/hP78B/dirnn+DbX0fB78Wyah79wG9ZT/f+coiigmfEfXBucznMqS+Nznk1voTtrsN/w/VbHVRQFU/oglLBYzFkjsY64GzU2Ff/BNWjx3VGjuwZf38Of4ctdjmXELOxTfoAppS/m7NHotaX4cpejN1WhRaeg2MIJC7PSkLcZ16fP4du/DMPnQUvpc05THwMVR3B9+jxqVBcct/8K/7GdBEryMPeeiKKoeDa/jb9gM4anBd+BlaAHQA/g2fIOns/eCL4me5fiL9hCoKoQo6UudH95dy/GvWYOvgOr8ZfsD97fqomWj/+IomrYp/0If/56MHRMqf2C7TlZgHv965gyRmAdcHOb9mqJPfAd3UrgxH6ihk6luaERX95KPJ/NReuajX3aj1FMpy+qDcEAkbnXBAyvC1/eSvz5G1DskThufrz1a2WynLFA9xnPoShnrX+mRHVBcURhGXAzyqlgjKKawNOML3998J60BKdc6vVl+PYuxdx/KpHpObi8BKfDWsPQkrJR47sRKNoJqhYKXHk2zQPAOuY+VHtEcOpiWCzWEbPatE1RNUw9huEvOxh8b2kmHNN/inoqsKhoZsw9hmLuOR5TWn+02NQ2NbRUR3Tw/ViwGb2pGjUmBVP6IExdc4JtjE0FVyP+4j34j23HX7gt+O/YDgKVhSjWMExpA7D0vRHL8K9jHfH1YLZa7nIMVyOROUNocflD5zMCPlyLfo939yL8J/ahaCb0ulJcn/4FvfIoWmwa/qNb8OWuwFewOfga3/ILTKn98R/fi//YTsy9J4Gi4tn8Fp7NbwEGlkG3BGt05a9HsdrRumSf02vu3f0x/oLN2CZ9F3PfG/AdWI1eX94qyHU6hmGc9v3akb7DLoZkQHVcl+s3mF3z07R7BfVRvdhepnLrmO6dov7h9fKePV/S785F+t25SL+vXZIB1clt3F9OYoydPt1jQo/5jmwM1qPJHBl6TOvaE7VLFt59n2DuPalVFohhGPj2L8Oz5V3AwH94I9bR92LKGg2Gjl5Xhqs5gOFIP+MPPc/mdzBa6rDf9syposhj8B1Yg+5uCmU1GJ5mvNvmoyXlYMocBQSzUOzTfkTzu7/Es+Ud7Dc9SqA0DwBTav9W5zBljMB/dGswsyo1WPNId9bSsuj3wdoxX3s6tAy9Kbk3geriYBHnRb/HdsP3gtk4RbuwjrkfS78pBCqP4s1bheFpxjz2m5i6DSZQWYBryZ9xLf0T9f3G4S/YjGXYnZjSWrfFnDmSQGUhvv3LUOPTsd/8YxSLA9uk79Dy4e9wb5qH/Ybvhbb3l+fjWf8v9IaKYDHrdopCm3PG4t35Ed69S0L9A9AbTuI/th3LgOntZgYpqglL70lfHCciDu++T/Bs/wAtfSD43Hi2vY/aJQvLwOmh4J1itmG76TG82+fj3b8M/+GNaKn9qHDYcR/ejhqXhpbaF9++TzAaT2Kb/L1QYKM9gepiWj59DiUsBvuMJ1Ad0dhG34dr+Qv48lah2KPw5a7A3O8mLINvxbP5bby7FgXbYovAMnAmWlIOgeoi9KpjBEpy8R/Z9MUJVA1TxnC0xEy8uxbR8sFvgzWEmmtx3PpLtC5ZmLLH4stbiaXfVNBMuFb+DSUsFtvYB9pts6KZg21c9hfK3/od7rIC8HvQUvpiv+nRc576pVgc2MbNxpQ1Gu/Oj4LX2XZpC/OftQ2KGsyy+gpznxvw7vsE775l2MbcB0Cg9AAApuQvMhS1uG5ocd1CfxstDXj3LAkWaTcMAmUHTxVhP3X/2MLbzZwLtcdsxXHz47jXvY4pa9Q5rTj4VdYx92EZNAMlLPa0nz+GHkBvPAmBwKkTK6iRie2+drabfoh323y8e5dy0u9Enfi9UOaWd8eHwanDA24mcHwf7rX/BEBN6IF95pNocd3QnTX4DqwmcPIotonfDmYwArax99Oy8L/wbHsfvbGKwIlgoNw68u7Q562vcBuenQsxZY1uk8lmuJ24Vr+CXlWEmtAdNTYVX+4KTJkjv8iiG/I1PFvfw398L6ZuA097zQLVxbiWv4h1zH2Yuw89j6stRMemOoKfqeGaFwMbLZ7goitCCCGE6BgkAHUdq6x3ceh4PXdMyPhiqoiu4y/YgpbWv9UUJEVRsA6ciWv5C8HsnOwxp7b34/lsLr5DazF1H4pl0AzcG+fhXvMq6u7F6M5a8HtoIRj8sU36TruDOv/xffjy15/KusoEwNxzAr7cFfiPbMLSfxqGoePe/A6Gx4l1zDdbDSbV8Dgsg2/Bu2MB/tID+E/sR7FFoMZ3a3UeU1r/4LSuwq2hAI1310II+HDc+mybrCEtPh3HHb/GtewF3MtfBMAy7M7QoFlLzMR+qr2hcyTlYJ/6GK5lf6Fu3dto3QZhGXxLu6+BdeQstNhUTOmDQ5klWnz3YF92LaTF3QSqCQJeAqUHUCLisU//aZtgVuh10sxY+k/Fs/U9ApWFoWw1795PQNUw929/2lCb46gmrEPvwL36lWDArqoIw9WEffpP2hRIV1QV68i7Mfefiu/gOnwH1+CqaAlONxwwDRQNX3x3PJvfomXh7zBljgxOeUro3ioYpteX41r6ZxSzHcfMn4UG2Fr6ILS0/nh2fARGAC0pB+uob6CoJuw3fA9/r4kYrobgNTx1b30+FdMwDIzmWgKVhRgtDZgyhoWOa84ajXvL28GA6bjZaF2ygq/J0NvwF2zGu/OjYKF1tzO4utsZgkGm9EGY0gfjKcnFlDkSS98pZyygfyampGxMM5+8oH0vFzUiHnPP8fjyVmHuMwktOvnU/ZiAGtl+PTUA66h78B/fg2fzO8FMOkXB3HP8eZ1bsYZhn/rYBbddUTWU8DPXq1NUDS363IoRK0rwflfCYmnZNBeTrmGb/Ah6VWEwQN9rArZR92CM/AaBsoMY7iZMPYaHpmCq4XFYR8xqc1ytSxamnLH48laBomId9yCWPpNbbWMbfR/N7z+NZ+v72Cd/N/S4Xl9Oy6d/wWiuwdRjOHptCb7SPBR7VKvAqbnfVHyH1uPeOBebqqGl9G0TlDPcTlwrXsJw1uDZOBdTSutFHoyAH09ZAd7DuejVxWip/c67Rpvh9+Ar2IJ+8iiW4XeecVqoEJeSdmoKXpjiAcDp8kkASgghhOhAJAB1Hdu0vxwFGNM3KfRYoOwgRks95pwxbbbX0geixqTg3bMUU9Yo9Loy3Ov/hV55FMugmViG3xWszXLbM/gOrMJftAtzSl+0xAzsehN1696mxVmDfeqPWtdU8jTjXv86akwKlqG3f3G+uDTUhAx8h9Zh7j0J95o5+I/tCAapTmUpfZllwM348jfg2TQXw+0MTgf6arDEZMGUPgjfsZ1Yx83GaKrGl78hWPPnK8Gnz6mOaBy3/gLPprdQIuKxDGo/mPRlprT+2G96FLV4G8rI+0+7qp2imtodkFsG34reVIVeW/rFYwNnYBly2xkziADMvSfj2f0xnq3vYc4Zi6EH8B3+LFj75TwGeqbMEah7luDZ+h5GSyPmXhPPuJqY6ojGOvQ2LINnEh/roKb+i/RQS7+bUCMT8Gx5D+/2D051XsGUPgRzvymoEQm0LPkTKEow+PSlgIGiKMGB9/xnUKz/P3t3HiVVea79/9o1jz03IjaCCEYUkVYkKnEewIADvEETo2JyPErUaIgQTXRFNGqixiEJglGTk+gPT8TEKOgvGjmiJvjmKJHkHIygoBgGkZ67qrrmvd8/iu7YNkM1dNfu6vp+1mKd1bt27bqfpSe0V93P/YTkO/Pqro4TST0Ghn+aYRgyQtXdntf1mi8k/6n/LuuEi7sFYY5wrdyHn6r0P/5LkuQ79d93+e/bZ/nOvEY1lT41tWf2em8x8hz3JaU/eEvJN56Uf+pcZT5e12N77Gc5QlU7Z2T9TnJ55Rw+vqvjp9h5xp2poNdS88olSro9yn78noxApbzHf0XSzq2tvRz47Z10oZTskPvIM7q2gH6ao/wAecZPVepvzytVVSfDF5KVTij512dz/9s7/aauINVKJyTL6jYvzHC65D35a0q8vFDx///HclQcKPeRZ8g9ZrIMjz83523lI7JizfKe8JVch+Ga5+Wd9CVJktneoI7nf6RotCn3QKdH6fW5UzQ7v5SQcoFY9pMNu1xjtmWr0uv/JCVjkgxltqyVf+q3unXOAf3FcPskp0t+5ea6RTvSUpXNRQEAgC4EUIOUaVl6Y+12jR1Zqeryf327nX7/Dcntl+vgngOwDcMhz9FfVOLVR5VYsUiZj9bIcPvlO32O3KOP/9d9Doc8486SZ9xZ3VjRuwAAIABJREFUXdcqa8OKe6qVeOURdTx7u3ynXyXX0MMkSYk3npQVb8/NyvnMfBv32FOUfP0/FPvt92W175D3+C/LfdSUXa7JcHnkPeErXZ1Kn91+18k1alJuPtDWd5V+b5XkdO01VDJcXvlO/toe7+nxOSPqVTvxZDU0RHr1Pin3H4r+067s9fuk3Dwhz1FnK/XXZ7uGF8vpkufoc3r3HMMh73H/R/GXHpQ8AXkm/Z/83udwyeH2Suq+P9l18AS5Dp4gKxlTtuHD3MD59X9SZtNfc11eLk/uBLOKoT2e6ag4UP4vzpcjWNHn3RK72pLoOebc3LD6MSfkto/l8xynSw6vX1Lv/3kXA4e/TN6JM5V8Y4lSq38vpeJyDtt7wOI5aorS616XFWmQ5/BT+7/QAqo4caaiLa1K/e0FSZL/i/N7PSD+0xyBcvmnXL/Hezz103NB4H8/9a/3VQ6Tf8rcbt1ou5v35TrwcwpefJ8yH7yl1DsrlFz1/yn55m/lPuwLkmEou/l/5P3CZfIccbqyjf/sGlwul0cdL9wjK53QkPO/pVigToa/TPEXH8htNXR75ao7KheO/e2F3Gy2XTEcco08Ru5xZ8lw+xR/6UF1LLtL/jPm7PLvHaAvGYYhwxuS14xLynVAAQCAgYMAahCyLEsf/fUNRdsjmnnyvwaKZ7a8o8wHb8k9+vjdzq1xjf68jNXPKPPharnGnCjvCV/ptlVvT9yHTJTjvBrFX/6Z4svuknvsqXIOPUyZ91fJc8x5u9yy5B41Scn/+5+yOlrlP/s6uUbW7/EzXCPq5awbp+yWtXJ+agZSt3vqxklun5JvPyfzkw3yTJi+22Pji5XnmPNz/9G4c4qr4fbt0zwh58FHyz1+qlxDP5f3P+e9MbxBuerGyVU3Tt6JM5XZ+N9Kf7ha3vpz99gFsadOp77mCFQoeMkD+zzwe7ByH3G60u++ptTfnpckOfPo8DFcHvlOvULpd1+V81OnVA4WnuO+JO0cDu/azf/m9CXD7VPwSz+QFW//17VgZbe5fHt9hssj92GT5T5scm6O3doVSr+7UjKzch32BbnH5rb+eT8/S5lNf1Xiz7+W1dEqKxFRYNp3FDryaMV3Buv+Kder44V7lFixWEaoWlb7J3KNPkHeY86TdvH/P4bH3y30DVzwfcVfelDxF38i58Hj5Rl31i63BgJ9xfCF5M4SQAEAMBARQA1C2U82qObtR/W98oCqvdWShir17qtK/vlxOSqHyTNxxm7fazhc8k/9tqxUh1xD8zuJ6dOctSMVnHWXkqufUXrtH5V+91U5qobLU3/erj/Pk5sHZHgCu+yM6XG/YeTmsTRu2m2nTG4bXr0yG/5vrrOnl51BxaBz61lfPMd3/Jf7oKLdPN/lkftzJ/V6LlAhED71ZDic8k7+quLP3y1HVZ0c/rK83uc68HNdp1EONoZhyHvM+YX9TJdHxj4MZN8V55BD5T/9UJnHf1nZre/IdcjErvDHEaiQ95jzc91WTrf853y7xymohtunwNRv57qjUh17nFG3K45gpQLnfk+pv7+g9Luv5ubAlQ+Vq26cnENGyTnkUDnKD+iTtQKSZHhDcmY6JBFAAQAw0BBADULJ7Rsl5cKdzCsPKfY/I3LDZIePl/+Mb+x1C4mz6qD9+nzD7ZXvhK/IPfp4pf7nRXnqz5Ph3P2/ap/9D569cfjL5Bi+504L96Gfz51Od/QXd7kFC8CuuYaNlWfiDDnKCAUGE0egXI4xPWf/ucedJTPSINfIY+QaNnaX7zV8IQVm3CrJ6Bq23huG2yvvxJny1J+rzAdvKb3+T7k/76yQJPmn37jbzwZ6y/CFZLRsk9NhKJYggAIAYCAhgBqE2jZvlEyf2k69STXt/63U28vkHneWvMd/uVfbOPaXs/YQ+c/4RsE+r9tnH3y0/GdfL+fB+X9TDyCn0B0/sI/hdMn3hcv2fl8f/N1hON1yjzlR7jEnyjJNma1bZbZul7P2kP1+NtDJ8IakZFRBn4sOKAAABhgCqMGoZbO2ZCs1blilvKOmyzP+i/v0rXUxMwxjr/OkAAD2MBwOOauGy1k13O5SMMgYvpCsBAEUAAADUWmlEiXAymYUSOxQi+sABXy5fLHUwicAAFCaDF9IskxV+y3FCKAAABhQSCYGGbN1m5wyZVbU2V0KAAAoIps2bdKkSZO0ZcuWXb5umqYeeOABnXzyyTr22GN14YUX6s033yxwlXvWeSJslTejCAEUAAADCgHUIBPb9oEkKTCUmRoAACA/K1eu1MUXX6y2trbd3vP0009rxYoVevrpp/XWW2/pi1/8oq666iolk8kCVrpnhjcXQFV40mzBAwBggCGAGmQiWzYqabk05OARdpcCAACKwOLFi/XjH/9Yc+fO3eN9GzdulGmasixLlmXJMAz5fD5ZllWgSveuswOq3JVWLJ4eULUBAFDqGEI+yJhN/9S2bIVGH1hudykAAKAIzJw5U3PmzNHWrVv3eN9FF12kFStW6JRTTpHT6ZTX69Wjjz4qn8/Xq8+rrg7tT7l7VHXgAeqQVBOwlMlaCpcH5PcO/l93a2vDdpdgC9ZdWlh3aWHdg9Pg/xu5hFiWpUB8uzY6D1XA57a7HAAAUAQOOOCAvO5LpVI65phj9Nhjj2nYsGH6xS9+oW9+85tatmyZamtr8/68pqaoTLPvO5Nqa8Nq7TAkSa50VJJXmzY3q6bc3+efNZDU1obV0BCxu4yCY92lhXWXFtZdvBwOY49fNLEFbxCxoo3yWEllyg6yuxQAADDIfOc739Fpp52mUaNGyefz6ZprrlE4HNaLL75od2n/4glIhiG/cnOpYvGMzQUBAIBOBFCDSGzbh5Ik7wEj7S0EAAAMOtu2bVM63X2wt8vlksfjsamingyHQ4YnKJ8VlyQGkQMAMIAQQA0ibZs3yLQMVY841O5SAADAIHPGGWdo8eLF2rx5s9LptH7961+rsbFRp556qt2ldecLyZPNBVCReMrmYgAAQCcCqEEk3fCRdphlOnhYtd2lAACAIrdt2zbV19dr9erVkqRbb71VJ510kr761a/qxBNP1IoVK/TLX/4y7xlShWL4QnJlOySxBQ8AgIGEIeSDiC+2TZuMIRrjZwA5AADonbq6Oq1fv77r52HDhmnNmjVdPweDQd1yyy265ZZb7Cgvb4Y3KEesRRJb8AAAGEjogBokrERUITOiZJgB5AAAoHQZvpCUiMrvdRFAAQAwgBBADRKxjz+QJHlqD7a5EgAAAPsY3pCsZFQhv0sxAigAAAYMAqhBou2Dd2RahspHHm53KQAAALYxfGEpk1K5z6ADCgCAAYQAapAwdrynrdlKHTCEAeQAAKB0Gd6gJKnKlyWAAgBgACGAGgSsTEqB6D+1IT1UFWGv3eUAAADYxvDlAqhKT4YACgCAAYRT8AaBbMOHclpZfew6SC4nmSIAAChdhicXQJW7M4rGDZurAQAAnUgrBoHsx+tkSYqERthdCgAAgK06O6DKXGklUlllsqbNFQEAAIkAalDIblunT6xqBcor7C4FAADAVp0dUCFnSpI4CQ8AgAGCAKrIWdm0sp9s1HupA1RdxvwnAABQ2jo7oAJGLoBiDhQAAAMDAVSRyzZ8KGVTei81RFVlPrvLAQAAsJfbJxmGfEpKIoACAGCgIIAqctlt62TJ0MbMAaoKE0ABAIDSZhgOGZ6gvFZnAJWxuSIAACARQBW97MfrlQoOVYflVXU5W/AAAADkDcptxiVJsQQdUAAADAQEUEXMymaU/eR9tQRyp9/RAQUAACAZ3qBc2YQktuABADBQEEAVMbPhQymT0jZnnVxOh8IBt90lAQAA2M7wBmSkYnK7HIp2EEABADAQFDSAamxs1Jw5c3Tsscdq8uTJWrhw4W7vveCCCzR+/HjV19d3/dm4caMkKRaL6Tvf+Y4mTZqkSZMm6fbbb1c6XXq/XGSbN0uSNqWrVV3mlWEYNlcEAABgP8MbkpXsUMjvpgMKAIABwlXID5s7d67q6uq0atUqbd68WVdeeaVGjhyp6dOnd7svlUppw4YNWrZsmUaNGtXjOT/4wQ/U1NSk//qv/1I0GtU3vvEN/fKXv9RVV11VqKUMCFYiIknaFnOoqsxjczUAAAADg+ENyEpGFfQRQAEAMFAUrAPqo48+0ptvvqkbbrhBPp9PY8aM0ezZs7VkyZIe97777rvyer065JBDeryWSCT0/PPP67rrrlM4HNaBBx6oa665ZpfPGeysRFTy+NUYyaiqjAHkAAAAUm4GlFIdCvmcijKEHACAAaFgAdT777+viooK1dTUdF0bNWqU1q9f3+PetWvXKhAI6LLLLtPnP/95zZw5U6+88ookadOmTUqn0xo9enS353zyySdqbW3t/4UMIFYiIsMbVms0yQByAACAnQxvULIsVfpMxeiAAgBgQCjYFrxYLCa/39/tmt/vVyKR6HGvZVk66qijNG/ePNXV1enll1/W9ddfryeeeEKZTKbrvZ9+jqRdPmt3qqtD+7KMvNXWhvv1+ZL0sRmXGSiTZUkjDqooyGfuid2fbxfWXVpYd2lh3UBxMrxBSVKlN6u1BFAAAAwIBQugAoFAj4AoHo8rFOoZBF1yySW65JJLun6eNm2ali1bppdeeknnnnuupFzYFAgEup4jaZfP2p2mpqhM0+r1OvJRWxtWQ0OkX579acn2VnUYuTV7HFZBPnN3CrXmgYZ1lxbWXVpYd3FzOIx+/7IJA9jOAKrMnVEsnpFpWXJwWAsAALYq2Ba8MWPGqKWlRc3NzV3XNm7c2G0rXaclS5botdde63YtmUzK6/Vq5MiRcrvd+uCDD7o9Z+jQob0KoAYDKxFV3MhtvWMLHgAAQE5nB1S5Ky3TshRPZmyuCAAAFCyAGjlypOrr63XPPfeoo6NDGzZs0OOPP64ZM2b0uHfHjh26/fbbtWnTJmUyGf3ud7/TmjVrNGPGDAUCAU2ZMkUPPPCAWltbtX37di1atGiXzxnsrHhEUTM3fJwh5AAAADmdAVTQkZIkTsIDAGAAKFgAJUk//elPFY1Gddppp+nyyy/XrFmzNGvWLG3btk319fVavXq1JOnaa6/V1KlTddlll2nixIl68skn9cgjj2jEiBGSpNtuu01DhgzRF7/4RZ1//vmaOHGirr322kIuxXZWJillU2rPehT0ueTzFGw3JQAAwIDWGUAFdgZQsTgdUAAA2K2gqcWQIUO0cOHCHteHDRumNWvWdP3sdrs1f/58zZ8/f5fPCYVC+uEPf9hvdRYDKxGVJDUn3aoqY/sdAABAp84AyqekpKBiCTqgAACwW0E7oNB3rERuQGxjwqVqAigAAIAuhssjOd3yWrkDcGJswQMAwHYEUEWqswPqkw6HKpn/BAAA0I3hDcqd3RlAJdiCBwCA3QigilRnB1QTHVAAAAA9GN6gXNm4JDqgAAAYCAigilRnB1TU8qoqTAcUAADApxneoJSMye91cgoeAAADAAFUkbISEVkyFLe8DCEHAAD4DMMblJWKKehzM4QcAIABgACqSFmJqDKugCwZdEABAAB8ljcgKxFT0O9mBhQAAAMAAVSRsuLtSjsDkqRwwGNzNQAAAAOL4Q3JSnUo5HMxAwoAgAGAAKpIWYmoEg6/nA5DHjf/GAEAAD7N8AakdEIhn0NROqAAALAdyUWRshJRxeVT0O+WYRh2lwMAADCgGN6gJKnCk6UDCgCAAYAAqkhZiYg6LK+CPpfdpQAAAAw4nQFUuTutWCIt07JsrggAgNJGAFWELMuSlYgqYnoV9LvtLgcAAGDA6Qygwq60LEtKJLM2VwQAQGkjgCpG6bhkZdWe8SjkI4ACAAD4rM4AKuTIbb+LJtiGBwCAnQigipCViEqSWtMeBdiCBwAA0ENnABV0pCSJOVAAANiMAKoIWYmIJKk55VKQDigAAICedgZQfiMhSYrRAQUAgK0IoIpQZwDVkvIo6KcDCgAA4LMMT0CS5LWSkqRYPGNnOQAAlDwCqCJkxXMBVNTy0gEFAACwC4bDKbn98ph0QAEAMBAQQBWhzhlQMdNHBxQAAOgTmzZt0qRJk7Rly5Zdvj5t2jTV19d3+/O5z31OP//5zwtcaf4MX1CubFwSM6AAALAb6UURshIRWQ6nkmIGFAAA2H8rV67UzTffrLa2tt3e88ILL3T7+cEHH9Srr76qSy65pL/L22eGJyilOuT1OBVLsAUPAAA70QFVhKxEVFlXUJJBAAUAAPbL4sWL9eMf/1hz587N+z1/+ctf9Otf/1o/+clPFAwG+7G6/WP4grKSMYV8LkXpgAIAwFZ0QBUhKxFR2pUbrMkWPAAAsD9mzpypOXPmaOvWrXndn81mdeutt+qaa67RiBEjev151dWhXr8nX7W14W4/fxIuV6rhnyoP+ZQ2rR6vDxaDdV17w7pLC+suLax7cCK9KEJWIqqUY2cARQcUAADYDwcccECv7l++fLni8bguvfTSffq8pqaoTNPap/fuSW1tWA0NkW7XUvIq0xGR1+1QS1uix+uDwa7WXQpYd2lh3aWFdRcvh8PY4xdNbMErQlYiooThkyEp4CVDBAAAhbN06VJdeOGF8nq9dpeyV4Y3ICsZU9Dn4hQ8AABsRgBVhKxEVB3yye91yeEw7C4HAACUiMbGRr399ts6//zz7S4lP96QZGZV7rU4BQ8AAJsRQBUZyzRlJWOKWl7mPwEAgIL629/+piFDhmj48OF2l5IXw5cbkF7uySiWyMiy+n7rHwAAyA8BVJGxklFJliJZL/OfAABAv9m2bZvq6+u1evXqrmtbtmzp9cwoOxme3MzMMldGWdNSIpW1uSIAAEoXLTRFxkpEJUltGY+CAQIoAADQN+rq6rR+/fqun4cNG6Y1a9Z0u+fyyy/X5ZdfXuDK9p3hyw1CDTuSkqRYIi0/8zMBALAFHVBFxkrkpuK3pD0K+vgFCgAAYHc6A6igIyVJisUzdpYDAEBJI4AqMp0dUM0JF1vwAAAA9sDw5gKogJGQJEU5CQ8AANsQQBWZzg6ohoSTIeQAAAB7YHhzQ8i9Vi6A4iQ8AADsQwBVZDo7oGImQ8gBAAD2xHB5JJdHHjMuSYol2IIHAIBdCKCKjJWMyXK6lRZb8AAAAPbG8IbkynRIogMKAAA7EUAVm1RclssvSWzBAwAA2AvDF5KR6pDH7VCMGVAAANiGAKrIWKkOZZ0+SaIDCgAAYC8Mb1BWMqqgz80peAAA2IgAqshYqQ5lnF5JUtBHBxQAAMCeGL6QlMgFUFG24AEAYBsCqCJjJTuUMnZ2QPnpgAIAANgTwxuSlYwp5HexBQ8AABsRQBUZK9WhpOGRRAcUAADA3hi+0M4teC5OwQMAwEYEUMUm1aGE5ZHH7ZDb5bS7GgAAgAHN8IYky1KF1+QUPAAAbEQAVUQsy5KV7FCH5WEAOQAAQB4MX1CSVO5OKZZIy7IsmysCAKA0EUAVk2xaMjOKZd1svwMAAMiD4Q1JkspdaWWyllJp0+aKAAAoTQRQRcRKdUiSIhkXHVAAAAB5MHy5ACrsTEkSg8gBALAJAVQR6Qqg0i5OwAMAAMhDZwdUwEhKkqLMgQIAwBYEUMUkmQugWlNOtuABAADkobMDyq+EJDGIHAAAmxBAFZHODqi2lJMteAAAAPnwBCQZ8inXARUhgAIAwBYEUEXE2tkB1Z52KeinAwoAAGBvDIdD8gbkNeOSpEgHARQAAHYgxSginR1QcctDBxQAAECeDG9IzszOWZodKZurAQCgNNEBVUS6BVAMIQcAAMiL4QtKyZiCPhdb8AAAsElBA6jGxkbNmTNHxx57rCZPnqyFCxfu9T2rVq3S2LFjtWXLlq5rK1as0NixY1VfX9/1Z/78+f1Z+sCQissynEqLIeQAAAD5MrwhWcmoQgGPomzBAwDAFgVNMebOnau6ujqtWrVKmzdv1pVXXqmRI0dq+vTpu7y/oaFBN954o0zT7HZ97dq1Ouecc3T//fcXouwBw0p2KOvySTLYggcAAJAnwxuU2bpN4YCbLXgAANikYB1QH330kd58803dcMMN8vl8GjNmjGbPnq0lS5bs8n7TNDVv3jxdeOGFPV5bu3atxo0b198lDzhWqkMZp0+S6IACAADIk+ELyUpEFfa72YIHAIBNChZAvf/++6qoqFBNTU3XtVGjRmn9+vW7vH/RokWqqKjYZQD1zjvvaNWqVTr99NN10kkn6ZZbblFbW1u/1T5QWKkOpY2dARQzoAAAAPJi+EJSOqEyv4NT8AAAsEnB2mhisZj8fn+3a36/X4lEose9b775pp577jk988wzisVi3V6LRCIaPXq0zjzzTJ1//vlqb2/XjTfeqPnz5+uRRx7Ju57q6tC+LSRPtbXhPn/mVjOpjMsnp8PQ8IMqZBhGn3/G/uiPNRcD1l1aWHdpYd3A4GB4c7/3VXmzinakZVnWgPs9CgCAwa5gAVQgEOgRNsXjcYVC3YOg5uZm3XTTTbr//vsVDod7BFDhcFhPPPFEt+fOmzdPs2bNUjQa7fG83Wlqiso0rX1czZ7V1obV0BDp8+emYxHFMuUK+d1qbIz2+fP3R3+teaBj3aWFdZcW1l3cHA6j379sQvEwfLl/FypcaZmWpY5khnmaAAAUWMG24I0ZM0YtLS1qbm7uurZx40aNHj26231/+tOf1NTUpCuuuEITJ07UtGnTJEnnnXeeli9frg8++EB33323stls13uSyaQcDoc8Hk9hFmMTK9mhDtOtUIBfmAAAAPJleIOSpDJ3bvsd2/AAACi8ggVQI0eOVH19ve655x51dHRow4YNevzxxzVjxoxu951//vn6+9//rtWrV2v16tV64YUXJEnLli3Tueeeq4qKCj399NNavHix0um0Pv74Y917772aMWPG4A+gUh2KZV0K8Y0dAABA3jo7oEKOpCRxEh4AADYoWAAlST/96U8VjUZ12mmn6fLLL9esWbM0a9Ysbdu2TfX19Vq9evVen1FVVaXHHntMb7zxho4//njNmDFD48aN0/e///0CrMA+VjYtZdNqz7jogAIAAOiFzhlQASMXQEXpgAIAoOAKNgNKkoYMGaKFCxf2uD5s2DCtWbNml+8ZOnRoj5PyJkyYoCeffLJfahyorFRcktSedinMCXgAAAB56+yA8lkJSW5F4gRQAAAUWkE7oLAfkh2SpNaUkw4oAACA3nB5JYdLHjP3hR5b8AAAKDwCqCJhpXIBVIfpUcg/uGddAQAA9CXDMGR4g3KmO+RxOxhCDgCADQigikRnABW3PGzBAwAA6CXDF5KVjCns9xBAAQBgAwKoImEl/xVAsQUPAACgd3IBVFThgFuROFvwAAAoNAKoItHVAWW6FaIDCgAAoFcMb0hWIqpwgA4oAADsQABVLNiCBwAAsM8MX1BWIqqQ360oQ8gBACg4AqgiYSU7ZMmhlFxswQMAAOglw7tzC57fpUicDigAAAqNAKpIWKkOpR1euZxOed1Ou8sBAAAoLt6QZGZV7rOUSptKprN2VwQAQEkhgCoSVqpDKcOrcMAtwzDsLgcAAKCoGL6gJKnCk5EkRdiGBwBAQRFAFQkr2aGEvAwgBwAA2AeGLyRJKnfmgicGkQMAUFgEUEXCSnUobnECHgAA6B+bNm3SpEmTtGXLlt3es2LFCl1wwQU69thjNW3aNL3yyisFrHD/GN5cABVy5AKoKHOgAAAoKAKoYpHqUMx0K8wAcgAA0MdWrlypiy++WG1tbbu95/XXX9d3vvMd3XjjjVq9erWuvfZaXXfdddq+fXsBK913Dn+ZJCmg3MnCbMEDAKCwCKCKhJXsUDTjogMKAAD0qcWLF+vHP/6x5s6du8f7nnjiCc2ePVsnnHCCDMPQOeeco6VLlyoUChWo0v1j7AygfGZnAEUHFAAAheSyuwDkx0rFFUmzBQ8AAPStmTNnas6cOdq6dese71u7dq2OP/54ff3rX9c777yjgw8+WPPmzet1AFVd3X+BVW1teLevWVZIUadLZe6UnI6AsjL2eH8xGSzr6C3WXVpYd2lh3YMTAVQRsMyMlEmqw/KohgAKAAD0oQMOOCCv+9ra2vT444/rZz/7mcaOHatnn31Wc+bM0bJlyzR8+PC8P6+pKSrTtPa13N2qrQ2roSGyx3sMX5liTY0K+Wu1oym61/uLQT7rHoxYd2lh3aWFdRcvh8PY4xdNbMErAlYqLkm5IeTMgAIAADbwer2aNWuWxo8fL7fbrVmzZunggw/W66+/bndpeTMC5bLibQoH3GzBAwCgwAigikEyN6sgbnkU9ntsLgYAAJSiQw89VMlkstu1bDZrUzX7xvCXyYq3KxzwEEABAFBgBFBFwEr9K4BiBhQAALDDRRddpCVLluitt95SNpvVU089pa1bt+rMM8+0u7S8OXYGUCG/m1PwAAAoMGZAFQFrZwdUwvIozBY8AABQANu2bdO0adP06KOPauLEiZo1a5YkacGCBdq2bZtGjhyphx9+OO8ZUgOB4S/PdUDVuuiAAgCgwAigisCnO6CCdEABAIB+UFdXp/Xr13f9PGzYMK1Zs6bbPbNmzeoKooqR4S+TLFNV3qw6khllsqZcTjYEAABQCPyNWwQ6A6iM0yuv22lzNQAAAMXJ8JdJkircCUlSLE4XFAAAhUIAVQx2BlBOX9DmQgAAAIqXESiXJJU5cgEU2/AAACgcAqgiYCU7ZMqQhwAKAABgn3V2QIWMzgCKQeQAABQKAVQRsJIxpeRRKOixuxQAAICi5fDnOqACVq67PMIWPAAACoYAqgiY0Sa1WUGFGUAOAACw77wByXDKZ8YkSW0xOqAAACgUAqgiYEUa1ZgNKUQABQAAsM8MwyHDH5Y7HZXTYagtSgAFAEChEEANcJZlyYw0qCETVChAAAUAALA/DH+5rHi7yoIetUWTdpcDAEDJyCuAeuWVV5TNZvu7FuyClYhImZSasyG24AEAAOw2gAR+AAAgAElEQVQnI1AmK96uipBHrQRQAAAUTF4B1A033KCTTz5ZP/zhD7Vu3br+rgmfYkUaJUlNZkihAEPIAQAA9ofh7wygvGplBhQAAAWTVwC1atUqzZs3T++9955mzpypCy64QL/61a/U3Nzc3/WVPHNnANVsMgMKAABgfzl2bsErD3qYAQUAQAHlFUAFAgHNmDFD//Ef/6GVK1fq3HPP1YsvvqhTTz1VV199tVasWMEWvX5iRhokSU3ZIAEUAADAfjL8ZZKZUY3fVDSeViZr2l0SAAAloddDyIPBoCoqKlRRUSFJ2rx5sxYsWKCzzz5ba9as6fMCS50VaVDG6VdSHgIoAACA/WT4yyRJVZ5c9xNdUAAAFIYrn5symYxeffVVLVu2TK+99pqCwaCmT5+u66+/XmPHjlUmk9Gtt96qG264Qa+88kp/11xSzEij4u5c2EcABQAAsH8Mf7kkqcKVG0DeGkuqutxnZ0kAAJSEvAKoyZMnKxaL6eSTT9Z9992nU089VS7Xv97qcrl00kkn6bXXXuu3QkuVGWlUxFkhn8cpt6vXDWsAAAD4FCOQ64Aqc8Ql0QEFAECh5BVAfeMb39B5552nqqqq3d5z+umna8qUKX1WGCTLMmVFG9XiqVNl2Gt3OQAAAEWvswMqqLgkr1qjSXsLAgCgROTVUnPppZfq8ccf15NPPtl1bebMmVq4cKEsy5IkeTweGYbRP1WWKKujTcpmtD0ZUE253+5yAAAAip7hDUmGIW82JsOQWumAAgCgIPIKoO6//3799re/1UEHHdR17aKLLtJTTz2lhQsX9ltxpc6MNEqSNnd4VcNsAgAAgP1mOBwyfGEp0a6ygEdtdEABAFAQeQVQy5cv13333adTTjml69pFF12kH/3oR3rmmWf6rbhSZ0UaJEnbEn4CKAAAgD5i+MtldrSrPORRW4wOKAAACiGvACoSiaimpqbH9QMPPFDNzc19XhRyOjugms0Qp7MAAAD0EcNfJiverooQM6AAACiUvAKoo446Sr/+9a+75j11WrJkiY444oh+KQySFWlUxhNWWi5mQAEAAPSRXADVpvKgh1PwAAAokLxOwZs3b55mz56tv/zlLzryyCMlSf/4xz/U0NCgxx57rF8LLGVmpEEJd4UksQUPAACgjxiB8lwH1EEetcdSypqmnI68vpcFAAD7KK+/acePH69ly5ZpypQpisfjSqfTmjp1qv7whz/omGOO6e8aS5YZaVS7USaPy6FwwG13OQAAAIOCw18mZdOqCliyJLXH0naXBADAoJdXB5QkDR8+XDfccEN/1oJPscysrGizmnyHqLrcJ8Mw7C4JAABgUDD85ZKkKndu+11bLKnKsNfOkgAAGPTyCqCSyaSeeuopvffee8pms13XU6mU1q5dq5deeqnfCixVVqxFsrLangowgBwAAKAPGf4ySVK5KyFJamUOFAAA/S6vAOq2227TCy+8oPHjx+uvf/2rJk6cqM2bN2v79u362te+1t81lqTOE/A2d/hUU8cAcgAAgL7SGUCFlAug2jgJDwCAfpfXDKiVK1fqRz/6kZ544gkNHz5ct956q1asWKGzzz5bHR0d/V1jSbIiDZKkrXEfA8gBAMBuWZal5557Ttu3b5ck/eIXv9D06dN1880383vabhiB3CEvATMqSZyEBwBAAeQVQEUiER199NGSpNGjR2vt2rVyOp266qqr9Prrr/drgaXKjDTKkqEWM0gABQAAdmvhwoVasGCBtm/frtWrV+u+++7TcccdpzVr1ujee++1u7wByfCHJYdTRrxVIb9brTECKAAA+lteAdSQIUP0ySefSJJGjhyp9evXS5LC4bCam5vz/rDGxkbNmTNHxx57rCZPnqyFCxfu9T2rVq3S2LFjtWXLlv16TrExI43KesuUlZMZUAAAYLd+//vf695779WECRP0hz/8QRMmTNCtt96qO++8Uy+//LLd5Q1IhuGQEaiQGWtRRcij1ghb8AAA6G95BVBnnXWWbrrpJq1Zs0Ynnniinn32Wa1YsUKLFi3S8OHD8/6wuXPnqrKyUqtWrdKvfvUr/e53v9Pzzz+/2/sbGhp04403yjTN/XpOMbIiDepw59rDa8qZAQUAAHatoaFB48aNkyT9+c9/1kknnSRJqq2tVTQatbO0Ac0IVsrqaFV5yKu2GAEUAAD9La8A6oYbbtAXvvAFbdmyRZMnT9bpp5+u6667TitWrNBNN92U1wd99NFHevPNN3XDDTfI5/NpzJgxmj17tpYsWbLL+03T1Lx583ThhRfu13OKlRlpVLtRJrfLobKA2+5yAADAADV8+HCtXbtW//jHP/TRRx/p5JNPlpSb4dmbLwpLjaOzAyro4RQ8AAAKIK9T8J555hldffXVqq6uliTdcccd+u53vyuv1yuXK69H6P3331dFRYVqamq6ro0aNaprO99nLVq0SBUVFbrwwgv10EMP7fNzdqe6OtSr+3urtja8z++1smlFYi1qDR+uA6oCGjKkrA8r6z/7s+ZixrpLC+suLawbxeCKK67Q3Llz5XA4dNxxx+nII4/UokWL9NBDD+muu+6yu7wBywhWytr8vyof4lF7LCXTsuQwDLvLAgBg0MorPbrvvvv0+c9/viuAkqRgMNirD4rFYvL7u28l8/v9SiQSPe5988039dxzz+mZZ55RLBbb5+fsSVNTVKZp9eo9+aqtDauhIbLP7zfbd0iytLXDp4qwZ7+eVSj7u+ZixbpLC+suLay7uDkcRr9/2TRQzJw5U0cccYS2bNnStf1uwoQJ+tWvfqXjjjvO5uoGLkewUsokVRWwlDUtReNplQU8dpcFAMCgldcWvLFjx+qNN97Yrw8KBAI9QqJ4PK5QqPsvh83Nzbrpppt07733Khzu+Q1svs8pZmZ7gyRpc4eP+U8AAGCvDj/8cJ155pnyer1qbm5We3u7hg4dandZA5oRrJQkVbtyv1e2sQ0PAIB+lVcHVHV1te644w49/PDDGj58uHy+7qey/fKXv9zrM8aMGaOWlhY1NzerqqpKkrRx40aNHj26231/+tOf1NTUpCuuuEKSZFm5LqXzzjtPt912m4466qi8nlPMzEgugNoS9+lznIAHAAD2YN26dbruuut055136vDDD9esWbO0detWud1uLV68WF/4whfsLnFAMgK5AKrC0SFJaosmNXzI4PlCEwCAgSavAMrn8+mCCy7Yrw8aOXKk6uvrdc899+j73/++tm3bpscff1xXX311t/vOP/98nX/++V0/b9++XaeccoqWLVumuro6ScrrOcXMijTKMhxqMwOqIYACAAB7cPfdd+uwww7ToYceqmeffVbxeFxvvPGGfvOb3+jBBx8kgNoNx84OqJBy4x5aopyEBwBAf8orgPrhD3/YJx/205/+VLfffrtOO+00ud1uXXzxxZo1a5a2bdumadOm6dFHH9XEiRP3+TmDhRlpVMZbIVMOVRNAAQCAPfjb3/6m3//+96qqqtLrr7+uU089VVVVVTrvvPP085//3O7yBqzOLXj+bERSmJPwAADoZ3kFUMuXL9/j6+eee25eHzZkyBAtXLiwx/Vhw4ZpzZo1u3zP0KFDe5xwt7vnDBZmtFEd7gpJYgYUAADYI4/HI8uylEql9NZbb+nOO++UlJur2dtDY0qJ4fJI3qAciTaF/FVqae/dgTYAAKB38gqg5s+fv8vrXq9XQ4cOzTuAQn6s9ga1u0bK7XKoLOC2uxwAADCATZo0Sffcc4/KysokSaeccorWrVunO++8UyeccILN1Q1sjkClrFiLqsKHqznCFjwAAPpTXgHUunXruv2czWa1adMmLViwQBdddFG/FFaqrExKVrxNTYGQqst8MgzD7pIAAMAAtmDBAi1YsEDr1q3T3XffrVAopOeee04+n0/f+9737C5vQDNClTJjLaoq86mxLW53OQAADGp5BVCf5XQ6deihh+qmm27St771LU2fPr2v6ypZZrRRkrQt6VdtBdvvAADAnlVXV+tnP/tZt2vz58+Xw+GwqaLi4QhUKtP4T1Ud4NV7m1vtLgcAgEFtv34zcTqd2rFjR1/VAklWey6A+mfUq9oKBpADAIC9e/nllzVr1ixNmDBBEydO1MUXX6w//vGPvXrGpk2bNGnSJG3ZsmW391xwwQUaP3686uvru/5s3Lhxf8u3jRGslBVvV3XYrY5kRvFkxu6SAAAYtPZ5CHk0GtXSpUs1fvz4Pi+qlHV2QG1N+HUUA8gBAMBe/OEPf9C3v/1tTZ06VRdccIFM09Tq1as1d+5cPfDAAzr77LP3+oyVK1fq5ptvVltb227vSaVS2rBhg5YtW6ZRo0b15RJskzsJz1KtN3cCXnMkqYO8+7RBAAAA7MU+DyF3uVyqr6/XggUL+rqmkma2N8hyuBSx2IIHAAD2btGiRfrWt76lq666quvapZdeqkceeUQPP/zwXgOoxYsX6/nnn9fcuXN1yy237Pa+d999V16vV4ccckif1W43RzB36nC1K3cCXkt7QgfVcHIgAAD9YZ+GkKP/WJEGpb0VsmSwBQ8AAOzVRx99pKlTp/a4PmXKFD300EN7ff/MmTM1Z84cbd26dY/3rV27VoFAQJdddpnee+89HXTQQbr22mt1+umn96re6upQr+7vjdracK/uT2brtFXSQWWmJCllGb1+xkBQjDX3BdZdWlh3aWHdg1PePcZPP/20AoGApk2bJkm69tprdcYZZ2jGjBn9VlwpMqNNijlz38bRAQUAAPbmwAMP1HvvvacRI0Z0u75u3TpVVlbu9f0HHHBAXp9jWZaOOuoozZs3T3V1dXr55Zd1/fXX64knntCECRPyrrepKSrTtPK+P1+1tWE1NER69R4z5cn93/YGGXLrn9ta1dBQ1ee19ad9WfdgwLpLC+suLay7eDkcxh6/aMprCPkvfvEL3XXXXcpk/jWY8dBDD9Xtt9+uJUuW7H+V6GK1N6hVYYX8bvmZQQAAAPbiS1/6km699VY9/fTT2rhxozZu3KilS5fqtttu69MvCi+55BItWrRIo0aNksfj0bRp03TiiSfqpZde6rPPKDTDF5YcLqmjVeUhj5rbk3aXBADAoJVXwvHkk0/q3nvv1Zlnntl1be7cuTriiCN033336atf/Wq/FVhKrFRcVjKqHUaA7XcAACAvX//61/XJJ5/otttuUzablWVZcrvd+vrXv65vfvObffY5S5YsUV1dnU455ZSua8lkUl6vt88+o9AMw5ARrJDV0aKqskPV1J6wuyQAAAatvDqgmpqaNGbMmB7Xx44dq+3bt/d5UaWq6wS8uI/tdwAAIC9Op1O33HKL/vKXv+ipp57Sc889p9WrV+uMM87QpZde2mefs2PHDt1+++3atGmTMpmMfve732nNmjVFP47BEaiUFWtRVdir5ggdUAAA9Je8AqjDDjtMy5Yt63H9hRdeGDTH8A4EVnsugPpn1EsABQAAeiUUCmn8+PH63Oc+J6/Xq7a2Nr399tv7/Lxt27apvr5eq1evlpSb/zl16lRddtllmjhxop588kk98sgjPWZPFRsjWCkz1qqqMp9a2hOyrL6fTwUAAPLcgnfNNdfoG9/4ht566y2NHz9eUu4klLfeeks/+9nP+rXAUtLZAbUjE9LJ5WzBAwAAhVNXV6f169d3/Txs2DCtWbOm62e326358+dr/vz5dpTXb4xgpayP/qaqQzxKZUzFEhmF/G67ywIAYNDJqwPqlFNO0ZIlS1RbW6vXXntNq1atUnV1tZ5++uleH72L3TPbG2Q6PYpZdEABAAAUgiNYIWVTqgnkOp+amQMFAEC/yPuYtfHjx+t73/ueqqurJUlvv/32LudCYd9Z0UYlPRWSDAIoAACAAjAClZKkancueGpqT+jgA8J2lgQAwKCUVwD14Ycf6oorrtDZZ5+tG2+8UVJuDkBZWZkeffRRDR8+vF+LLBVmpFFRR5kchqGqsuI9UQYAAPSvhx9+eK/3bNq0qf8LGQSMYC6AKnd0SJKa2xlEDgBAf8grgLrjjjt05JFH6qqrruq69sc//lHf/e53ddddd2nx4sX9VmApsaLNanUcqqoyr5yOvHZHAgCAErR06dK87jvwwAP7uZLi59gZQPnNiJwOQ80RtuABANAf8gqg1qxZo2eeeUYVFRVd10KhkL71rW/py1/+cr8VV0qsTFJWMqoGl5/tdwAAYI9eeeUVu0sYNDo7oBRrVWW4Vi10QAEA0C/yarPx+/3asWNHj+stLS1y0KnTJ6xosyRpWwcDyAEAAArFcLpl+MtkxZpUFfYyhBwAgH6SV3p09tlna8GCBVq9erWSyaSSyaRWr16t2267TWeccUZ/11gSzM4AKuFVbYXP5moAAABKhxGqlhltVlW5T010QAEA0C/y2oI3b948XX/99brkkktkGIYkybIsnXnmmfre977XrwWWCivaJElqMYN0QAEAABSQI1gls/VjVQ3xqTW6Q6ZpyeEw7C4LAIBBJa8AKhgM6rHHHtMHH3yg999/Xy6XS7W1tfr73/+ur3zlK1q+fHl/1znombFmWTLUZgYIoAAAAArICFXJ3PqOqsIeZU1LbbGUKsOcSAwAQF/KK4DqNGrUKLW2tmrp0qV68cUXlUgkdPjhh/dXbSXFijYp7QopKycBFAAAQAE5QlVSOqEavyVJao4kCKAAAOhjeQVQkUhEzz77rJYuXaoNGzZIkiZPnqwrrrhCxx9/fL8WWCrMaLNizrB8HqeCvl7lggAAANgPRqhaklTt6pCk3El4w+ysCACAwWePScdf//pXLV26VC+99JISiYSOOOIIffvb39aDDz6om266SaNHjy5UnYOeFWtWqxlSbYW/a84WAAAA+p8jWCVJKjNikqQmTsIDAKDP7TaAmj59ujZu3KixY8dqzpw5OuecczRixAhJ0oMPPliwAkuBZVkyo01qytaqupIT8AAAAAqpswPKm2qVx+1UMyfhAQDQ5xy7e+GDDz7QiBEjdNppp2nixIld4RP6QTImZVL6JOlTZRnzBgAAAArJ8JdLhlNWrEXVZT46oAAA6Ae77YB6/fXX9dxzz+nZZ5/VokWLVF1dralTp2rKlClsEetjZrRJkvRJ0q9DQwRQAAAAhWQ4HDKCFTKjTaouH63GtrjdJQEAMOjstgOqpqZG//Zv/6bly5frqaee0llnnaXly5frsssuUzab1W9+8xt9/PHHhax10LJizZKkFjPIiSsAAAA2cISqZcWaVVPuV1MbHVAAAPS13QZQnzZ+/Hjdeuut+vOf/6z7779fJ598sv7zP/9TZ555pq699tr+rnHQM6MEUAAAAHYyQlUyo82qLvMqlsgonszYXRIAAIPKHk/B+yy3261zzjlH55xzjhobG/Xss8/queee66/aSoYVbZJpOBW1fARQAAAANnCEqpX54C3V7JzH2dSeUF1tyOaqAAAYPPLqgNqVmpoaXXHFFVq+fHlf1lOSzFizUu4yWTIIoAAAAGxgBKskM6taX1qS2IYHAEAf2+cACn3HijYr6gjL73XK5+lVUxoAAAD6gCNULUmqcnZIkhoJoAAA6FMEUAOAGW1SmxVUZdhndykAAAAlyQhVSZIC2Xa5nIaa2gmgAADoSwRQNrNMU1asVU2ZgCpDHrvLAQAAKEmOYC6AUqxZVWU+tuABANDHCKBsZsXbJCurHSkfHVAAAAB28QYll1dmtFk15T624AEA0McIoGxmRZskSR93+FTBAHIAAABbGIYhR6hKVqxZ1WU+tuABANDHCKBsZkabJUlNZoAT8AAAAGxkhKplRptUU+5TeyylVDprd0kAAAwaBFA2s2K5DqhWM0gABQAAYCNHsEpWtFnV5bmxCHRBAQDQdwigbGZGm5V1ehW33KoMEUABAADYxQhVy4q3qSbklkQABQBAXyKAspkVbVbSXSbJUGUZARQAAIBdHKHcSXjV7rgkMYgcAIA+RABlMzPWrJgjLJfTUNjvtrscAACAkmWEqiVJYSsih2GoiQAKAIA+QwBlI8s0ZbZ+rBaVqyLklWEYdpcEAABQshzBXAeU0dGiqjIvARQAAH2IAMpGZuvHUjqhzdkaBpADAADYzAhVSsp1qFeX+dTIDCgAAPoMAZSNzB0bJUnvJaoIoAAAAGxmuLwyfGFZkSbVlPvogAIAoA8RQNkou+MDyRPQhxEfARQAAMAAYIRrZEYaVF3uU2skqUzWtLskAAAGhYIGUI2NjZozZ46OPfZYTZ48WQsXLtzlfdFoVDfeeKOOP/54HXPMMZo9e7befffdrtfXrVunww8/XPX19V1/vvrVrxZqGX0m27BRqh6pZMZSZYgACgAAwG6OcK3MaKOqy32yJDVHknaXBADAoOAq5IfNnTtXdXV1WrVqlTZv3qwrr7xSI0eO1PTp07vd94Mf/ECNjY166aWXFAwG9ZOf/ERXX321Vq5cKUn63//9Xx199NF66qmnCll+n7LSSZnNW5QYM0WSVEEHFAAAgO0c4RplNr2tmrBHktTUGteQCr/NVQEAUPwKFkB99NFHevPNN/XAAw/I5/NpzJgxmj17tpYsWdIjgLrjjjtkmqa8Xq/a2trU3t6uqqqqrtfXrl2rcePGFar0fpFt3CRZltr8B0lKqyrss7skAACAkmeEayQzo2pvSpIYRA4AQB8pWAD1/vvvq6KiQjU1NV3XRo0apfXr1/e41+12S5IWLlyohQsXKhgM6uGHH+56fe3atfL7/ZoyZYoikYiOO+44ffe739XQoUPzrqe6OrQfq9m72trwHl9v3bBFcUmxshGSNujQEVWqrQr0a039bW9rHqxYd2lh3aWFdQOlxxGulSSVW+0yJAaRAwDQRwoWQMViMfn93duX/X6/Eond/6V+xRVX6Morr9SSJUv07//+71q2bJkOPvhgVVdX6+ijj9all16qbDarO+64Q1deeaV+//vfy+l05lVPU1NUpmnt15p2p7Y2rIaGyB7viX/4roxwrTY1ZWVIyqbSe33PQJbPmgcj1l1aWHdpYd3FzeEw+v3LJgxOjnDuy1JHrEkVYa8aCaAAAOgTBRtCHggEeoRN8XhcodDufzn0+XzyeDz62te+poMOOkivvPKKJOmRRx7RNddco7KyMlVWVurmm2/W+vXrtXHjxn5dw76yLFOJ13+l7Ccbuq5ld3wg55BRaokkFQ565HJyICEAAIDdjFC1JMmMNKq23KeG1rjNFQEAMDgULPUYM2aMWlpa1Nzc3HVt48aNGj16dI97Z8+ereXLl3e7lkqlVFZWpqamJt19991qbW3t9pqUC6wGIqu9Qel1ryr+X4tlpeIyO1plxZrlrM0FUJyABwAAMDAYLo+MQEUugKr0awcBFAAAfaJgAdTIkSNVX1+ve+65Rx0dHdqwYYMef/xxzZgxo8e9EyZM0MKFC7VlyxalUik99NBDSiQSOv3001VeXq6XXnpJ999/vxKJhFpaWnTbbbfphBNO0P9j787j47rre/+/zjmzb5qRNJIsyXu8JSGOk2AS0izAJaGkJTT9hf5qSin8Upof0CU3bUhb2l5CWnjQ/grlYQht0pampb300hRIWpp7S1wIgSQNdWJnc7zHsmzty+zLOd/fH0dW4ni3NRpLej8fDz9sjc6c+XxlWZ55z+f7OUuWLJmt5ZwRb/wQACY/QuXJf8Ad9Du1nI4VjEyWyegKeCIiItJk+/btY+PGjfT19Z3y2CeeeIJ169ad1rFzkZVsx+SG6EhHmchXqdTcZpckIiIy583qvq8vfvGL5PN53va2t/FLv/RL3Hrrrdx666309/ezYcMGnnnmGQA+/vGPc8MNN7Bp0yauueYannvuOR588EHS6TSBQID777+f/v5+rr32Wm644QYSiQRf+MIXZnMpZ+RIABVcez21l79P9bnvgOXwwkSC/uECa5dmmlyhiIiILGRbtmxh06ZNTExMnPLYoaEhPvGJT+B53ixU1hx2MouXGyKb8eeXahueiIjIuZu1IeQAHR0dbN68+Zjbu7u72bp16/THwWCQO++8kzvvvPO451m5ciUPPPBAw+qcad54P1Y0Rfitm3AHXsEb2IXVtpS//499dLXGePtlPc0uUURERBao++67j0ceeYQ77riDT37ykyc91vM8fvM3f5P3ve99fOlLX5qlCmefnWynvvspOlJ+l/rQWInerIbai4iInItZDaAWKnf8EHZ6EVYgROT6X6b4zU/TRyeDYyX++8+t1wByERERaZpbbrmF22+/nYMHD57y2C9/+cuk0+lzCqAaeXXCbDY5I+eZ7O5leKvH6k7/OVqx7s3YuRvhfK6tkbTuhUXrXli07vlJAVSDGWPwxg8RXLERACe7nPoNd3P/N/ayYVU7Fy9va3KFIiIispB1dnae1nFPP/003/rWt3jooYcoFApn/XgjI3k8z5z1/U8km00yNJSbkXPV8V8AlAb6iIYD7Okbn7Fzz7SZXPdconUvLFr3wqJ1z122bZ30jSa13jSYKeegUsBOL5q+7X89VyPnhfm5d6xqYmUiIiIip2d0dJS7776bP/7jPyaZnN/vzoK/BQ+A/Agd6ShDY5oBJSIicq4UQDXYkQHkRwKoofEST744wA1vXkxHOtrM0kREREROy+OPP87IyAi33XYbV1xxBTfddBMA73nPe3j44YebXN3MsxKtYFnTg8gHNYRcRETknGkLXoO9MYB6fNshLAsNHhcREZE54+abb+bmm2+e/vjw4cNcd911fPvb36a3t7eJlTWGZQew4q14uWE60pey9ZUhXM/DsfXerYiIyNnS/6IN5o0fgkAIK9GK63n8YFs/b1rRRmsq0uzSRERERE6ov7+fDRs28MwzzzS7lKawk+2Y3DAdmSiuZxidrDS7JBERkTlNHVAN5o0fwm5ZhGXZbN89zHi+yvvf2d3sskRERESO0tvby44dO6Y/7u7uZuvWrcc9tqur66hj5xKyuj0AACAASURBVCMrmcXte356ZMLgeImsxieIiIicNXVANZg3fmh6+933n+0nFQ+x/gJd+U5ERETkfGYn2zHFcTpS/vu1GkQuIiJybhRANZCpVzG5Yez0IsZyFbbtHuHqN3URcPRlFxERETmf2cksACnyBBxbg8hFRETOkZKQBvImBgCDnV7EE9sP4RnDtZdo+52IiIjI+c5Ktvt/yA+TTUfUASUiInKOFEA1kDfeD4DV0sXj2/pZuyRNZ2usyVWJiIiIyKnYUwGUlxsmm46qA0pEROQcKYBqIG/8EGCRD7YyNF7mstXZZpckIiIiIqfBimXAdjC5ITqmAihjTLPLEhERmbMUQDWQN34IK9nOeMkDoC0VaXJFIiIiInI6LNvGSrTj5YbIZqJUqi65Yq3ZZYmIiMxZCqAa6MgV8MZzVQDSyXCTKxIRERGR02W3dOBNDNKRjgJoG56IiMg5UADVIMZ4eOOH/QAqXwEgnVAAJSIiIjJX2KlOvMkBOtJ+F7sGkYuIiJw9BVANYgpj4FanAyjLglQ82OyyREREROQ02S2dUCvTFq5hAQNjxWaXJCIiMmcpgGoQU84BYEVTjOUqpOIhHFtfbhEREZG5wk51+r8XhsikwgxpC56IiMhZUyLSIKbqP0GxQlHG81VtvxMRERGZY+yWDgDMxIB/JTxtwRMRETlrCqAapVoGjgRQFTIKoERERETmFCvZDpaNNzFAV2uMw6PagiciInK2FEA1iKn6T1CsoB9ApROhJlckIiIiImfCsgNYySze5ACdrTEK5Tr5Uq3ZZYmIiMxJCqAaxNT8Fu26HSZXrJFOqgNKREREZK6xWzrwJgbpao0BcHhEXVAiIiJnQwFUg5ipLXiTNf9LrBlQIiIiInOPnerEmxygKxMF0DY8ERGRs6QAqlGqRbADjJcMoABKREREZC6yWzqhVqYtXMWxLQVQIiIiZ0kBVIOYWtkfQJ6rAGgGlIiIiMgcZKc6/T/kh8imowwogBIRETkrCqAaxFRLMHUFPICMZkCJiIiIzDl2ix9AGV0JT0RE5JwogGoQUy1iBaOM5Ss4tkUiGmx2SSIiIiJyhqxkG1g23sQAXW0xBsZKeJ5pdlkiIiJzjgKoRpneglclnQhjWVazKxIRERGRM2TZAaxk1h9E3hqj7nqMTJabXZaIiMicowCqQUy16AdQ+QrppOY/iYiIiMxVdkun3wHVGgPQHCgREZGzoACqQUy1PD0DSlfAExEREZm77FQH3uQgnZkoAIcUQImIiJwxBVCNUi1hBaOM56tkFECJiIiIzFl2SyfUyiSsEtFwQB1QIiIiZ0EBVAMYYzDVEq4TplSpk9YV8ERERETmrOkr4U0O0NUa1ZXwREREzoICqAYw9SoYl7LxZz+lE5oBJSIiIjJX2akjAdQgXa0xdUCJiIicBQVQDeBVSgAUvQCAZkCJiIiIzGFWsh0sB29igM7WGCOTFSo1t9lliYiIzCkKoBrAq/jviuXrDgAZbcETERERmbMs28FKtuNNHJ6+Et7gWKnJVYmIiMwtCqAawEwFUJM1dUCJiIiIzAd2uuuoAEpzoERERM6MAqgGONIBNVGxCQcdIiGnyRWJiIiIyLlwMj1444fpaPHfWFQAJSIicmYUQDXAkRlQY2WLdCKEZVlNrkhEREREzoWd6QGvTrA8TCYZ5vCIAigREZEzoQCqAbyq/4RkuKjtdyIiIiLzgZ3pBsAbPUhXa0wdUCIiImdIAVQDHNmCN1y0NIBcREREZB6w01MB1Fg/3W1x+kcKGGOaXJWIiMjcoQCqAY5swTucN+qAEhEREZkHrGAYK5nFGztITzZOpeoyMlFudlkiIiJzhgKoBvAqRXBCVOqQToSaXY6IiIiIzAA70403dpDebAKAvuFCkysSERGZOxRANYBXKeIF/M6nZFwBlIiIiMh84GR68CYO093mP887OJRvckUiIiJzhwKoBvAqRUwgAkAo4DS5GhERERGZCf6V8FzC5RHaUmH6htQBJSIicroUQDWAVynhTQVQwYC+xCIiIiLzgd3aAzA1ByqhDigREZEzoHSkAbxKEc9RACUiIiIyn9jpRYCFN+rPgTo0UqTues0uS0REZE5QOtIAplrEdfzZAAqgREREROYHKxDGSr12JTzXMxweLTa7LBERkTlhVtOR4eFhbr/9di6//HKuvvpqNm/efNzj8vk8n/jEJ7jyyiu57LLL+OAHP8hLL700/flCocBdd93Fxo0b2bhxI/fccw+1Wm22lnFKXrlI3Z4KoBwFUCIiIiLzhZPpwRvvn74S3kHNgRIRETkts5qO3HHHHWQyGZ544gm++tWv8k//9E888sgjxxz36U9/muHhYR599FGefvppLrnkEj760Y8e9fmRkRG++93v8q1vfYv/+q//4q/+6q9mcykn5VVL1NUBJSIiInPIvn372LhxI319fcf9/KneIFwo7EwP3vgAnS0hbMuiT3OgRERETsuspSP79+/n6aef5s477yQSibBq1So++MEP8rWvfe2YY++9916+/OUv09LSQqFQYHJyktbWVgDK5TKPPPIIv/Zrv0YymWTRokV87GMfO+55msEYg1cpUbMUQImIiMjcsGXLFjZt2sTExMQJjznVG4QLhZ3pBuPiFIboaoupA0pEROQ0BWbrgXbu3Ek6naa9vX36thUrVrBjx45jjg0GgwBs3ryZzZs3E4/H+cpXvgL4787VajUuuOCCo84zMDDA+Pg46XS6wSs5hXoFjEfNVgAlIiIi57/77ruPRx55hDvuuINPfvKTJzzu3nvvxfM8wuEwExMTR71BuJDYmdeuhNebjbOnf7LJFYmIiMwNsxZAFQoFotHoUbdFo1HK5fIJ73PbbbfxkY98hK997Wv88i//Mt/+9rfJ5/PT9339eYCTnuuN2toSZ1L+aavnauQBKxwDoKszRTIWashjnW+y2WSzS2gKrXth0boXFq1bFoJbbrmF22+/nYMHD570uJO9QXgmGvUcDGbne9fLrGafZROpDLF66QaefmmQeDJCLBJs+GOfyEL9N6t1Lyxa98Kidc9PsxZAxWKxYwKiUqlEInHiJyGRSASAD33oQ3zjG9/gscceY+PGjYAfNsVisenzACc91xuNjOTxPHNGazgd7vgQALmK3/k0MV6kXKjM+OOcb7LZJENDuWaXMeu07oVF615YtO65zbathgYd80lnZ+cZHX+8NwiXLFly2vdv1HOw2fzetVJZ8gf3klnqPy/d9vIAK3taZuWx32i+/Js9U1r3wqJ1Lyxa99x1qudfs7Y/bNWqVYyNjTE6Ojp92+7du4/aSnfEBz/4QR5++OGjbqtWq6RSKZYtW0YwGGTPnj1Hnaerq+uMAqiGqfphWAW/60lXwRMREZH5JBKJEAqF+NCHPkRPTw+PPfZYs0uadU6mF3e0j56OqSvhDWsOlIiIyKnMWjqybNkyNmzYwOc+9zmKxSK7du3iwQcf5Gd+5meOOfbSSy9l8+bN9PX1Ua1W+dKXvkS5XObtb387sViMG2+8kc9//vOMj49z+PBhvvzlLx/3PM1gpgOoII5tYdtWkysSEREROXcne4NwobGzyzATh2mLuISDDn2DuhKeiIjIqcxqe84Xv/hF8vk8b3vb2/ilX/olbr31Vm699Vb6+/vZsGEDzzzzDAAf//jHueGGG9i0aRPXXHMNzz33HA8++OD0gPFPfepTdHR08O53v5ubb76ZK664go9//OOzuZQTOhJAlU1QA8hFRERk3jjZG4QLjdOxEgAztJfu9jh9QwqgRERETmXWZkABdHR0sHnz5mNu7+7uZuvWrdMfB4NB7rzzTu68887jnieRSPCZz3ymYXWek6kAqmRCBAPzf/aTiIiIzE/9/f3cdNNN3H///dNv9nmex6ZNm6hUKqxfv/6oNwgXEie7HLBwB3fTm13D1p3DGGOwLHW+i4iInMisBlALgalNBVBekGCg1uRqRERERE5Pb28vO3bsmP74TN8gXEisUBQ704M7uJvl3W/m8W2HGBwv0ZmJNbs0ERGR85b2iM2wI1vwSl5AA8hFRERE5imncwXu4B4u6PZnYO3qm2hyRSIiIuc3JSQzzFRLWMEIVRfNgBIRERGZp+yOlVAp0BXMEw0H2HVQAZSIiMjJKCGZabUSdjhKre4pgBIRERGZp14bRL6HlT0pBVAiIiKnoIRkhpnqkQDK1RY8ERERkXnKTndDMII7uJsLelroHypQLGv+p4iIyIkoIZlhfgAVp+4aAuqAEhEREZmXLNvG6ViBO7ibVT0tGGB3/2SzyxIRETlvKSGZYa91QHnqgBIRERGZx5zsCryRAyzrCGNblgaRi4iInIQSkplWK2GFotRczYASERERmc+czpVgPIITfSzuSGgOlIiIyEkoIZlhplrGDsf9GVAKoERERETmLXtqELk3NQdqT/8kruc1uSoREZHzkxKSGWaqRV0FT0RERGQBsKMprGQWd3APK3tTVGoufYOFZpclIiJyXlJCMoOM8aBWxg7HpmZAOc0uSUREREQayOlYiTuwi1XdLQDahiciInICCqBmUq0C4AdQmgElIiIiMu85PeswxXFa6kNkkmEFUCIiIieghGQGmWoRACsUoVb3CCiAEhEREZnXAosvAcA9sI0LelrY1Tfe5IpERETOT0pIZpCplv0/hGIYgzqgREREROY5O57BbluK++pzrF6cZmSywsBosdlliYiInHeUkMykWgkAzwkDEHT05RURERGZ7wJLLsEd3MX6Xv854LO7hptckYiIyPlHCckMstuWEFr/bqyuNYA6oEREREQWgsCS9WAMLbld9GYTbN2pAEpEROSNlJDMICsQIvyW91G3goACKBEREZGFwM6uwIokqb/6HJeuamdn3zj5Uq3ZZYmIiJxXlJA0QLXuAtqCJyIiIrIQWLaNs/gS6ge2s+GCVoyBbbvVBSUiIvJ6SkgaoFbzAHVAiYiIiCwUgSXroVKg1xoknQjxrLbhiYiIHEUJSQMc6YAKKIASERERWRACvReBZeMd2MalF7Szfe8otbrX7LJERETOG0pIGqCqDigRERGRBcUKx3G6Vk3PgapUXV5+dazZZYmIiJw3lJA0QE0zoEREREQWnMDSDXijB1iTzBMOOtqGJyIi8jpKSBqgWlcHlIiIiMhCE1xzDQTCeC/8by5e3sqzu4YxxjS7LBERkfOCEpIG0BByERERkYXHCscJrr2W+q6nePPSIGO5Ci+/Ot7sskRERM4LSkga4MgQcgVQIiIiIgtL6E03AIaLK1tJRIP8+zMHml2SiIjIeUEJSQNMDyHXDCgRERGRBcVOZgmseDPuju/x3y5p49mdwwyOFZtdloiISNMpIWmAmjqgRERERBas0CU/CbUy18Z3Y9sW3/3xwWaXJCIi0nRKSBrgSAdUQB1QIiIiIguOk12G072OwM7HeMvaVh7f1k+pUm92WSIiIk2lhKQBaq46oEREREQWstBl78EUxnhP5BnKVZcfbDvU7JJERESaSglJA9RqHpYFjm01uxQRERERaYJA9zqCl7yL6L7HubFriH//8QE8zzS7LBERkaZRANUA1bpHMGBjWQqgRERERBaq8Jv/L+zsCm50t+BNDvGdp/Y3uyQREZGmUQDVALWaqyvgiYiIiCxwlhMg+o7/F8ex+Gj7j/j293by/J6RZpclIiLSFEpJGuBIB5SIiIiILGx2Kkvkuttorx/mV1u38NfffpbB8VKzyxIREZl1SkkaoFp3FUCJiIiICADB5ZcTedtHWGId5rbwd/jLbzxNpeo2uywREZFZpZSkAWo1j2DAaXYZIiIiInKeCK56K7EbfpXu4Di31h7i2//6g2aXJCIiMqsUQDVAta4ZUCIiIiJytMDSDcRv+i1aQ1VuGPgqe7/1FUw53+yyRERkATFeHVOvNOWxA0151HnO74BSACUiIiIiRwssWkPLz3+Op75+PxcdforJf9hGeNVVBBa/Cad7HVYw3OwSRUQWJFMpgB2AQGjGrmhvjAeAZb2WD5hyntorP6De/xKB3osJrHwLdjSFV6tQ2/dj3Fe3+wGRZYFlYUdbsFId2KkOrGgLViiCFYxg3BqmlMOUJjDVItRrGLcGxmCFov4xxsNMDuK97pfJDYMTIPGLm7ECoRlZ5+lSANUA1bpLwJmZb1gRERERmV+cWJILb/0YX/7qo7zT2sqaVx6n9uJ3wQ7gLFrth1G9l2DHWjD1CqZWwY5nsELRZpcuIg1ijMEb3o+d6Z71UOB8ciSwOaP7eC4mN4Q3fghTr2Ins1jJdgDcQztw+1/Gyw1ht3ThtPZitXSC50K9ginlcAd2UT+8AzMx4J/QcrDCMaxUB05rL3amB1Mt4o324Y31QyCEnenGzvRghaKYagmqJUxpAi83jJcbxpQmwHXBuGAHsNOLsFt7wLKp7/lPcGtY8VYqrz5H5Uf/E6dzJfmRVzG1MoSiWOGEX4vnUi9N+PWei1AMO9WB074Me8VGnM6VTfk+UwDVANW6RzSoGVAiIiIicnzpRJibf+on+P++nqC3NcJHrgrRVtiFe2A7lSe/Dnz96Ds4IQIrriC45hqcRWuPeXfeeB54NazAyTuojFvHcvQSQOYeYzzcgV3YkRR2uqvZ5cwoLz9K+QcP4r76LHa6m8j1t+F0rDjr85lKAVMpYCXasewz35ljKgWwbAhGsCzL76Ip5zH5UbzCCCY/himMgmVjJVqxE+3gBDDFcUxxAuNWsSJJ/1c47nd2ToUdJjfsd+IUxrECQQhEsGwLd/Qg3vA+vLFDVNp7oGsdgZ4LMdUS7vB+vNED4Nb9mkIRv9unNIlXmsDkR8Grn3hBgTB2qoNa/8vU3Oqxnw/HCXStxl5zDWBDtYip5PHGD1Hb+wy8/D3A8gOpTDemXsXtf4n6zh++dg7LwYomsZLtfrgTbcFygv7XpVbBGzuIe3gnplIguPpqghe+HadtCe5oH7VXnsA9+AKJi6+lvmg9TvcaLPu1n9PG8zCFUb97qZKfDrxwAv7jRFP+1zkQAifo36dWgmoZwO+ciiTO+PugEfS/TwPU6x7BaLDZZYiIiIjIeezCZa389/ddyl/+y4v8j+/keO81V/KTP/tzUBzF7XvBfyc8GMZygriHd1Lb9ST1nT/EirbgdK3C6VrNZCZF6eUfUz/4ItTKBFa82X9h07lqOqQyxuAeepnq1kdwD75AcM01hK/apI4qOStmqhPDsmfuDXdTr2KK48cNTEw5z/iTWyg8829+h4plE7z4nYQvf+8Jv4dNrYyXH8UURjGVov/iPOqHIThBPxiwHb+rxHMxXt0PUyzbD7oOv4J7YBv1gy9hBSPY7Utx2pdiJ9shGPUf17KgXp2u3Rs/5HfguHUC3Wtwui/ESrTi9r+E2/cC3vgh7I4VBLrXYnes9O9bmsQd2EXlPx8CzyW0/t3Udj1J8Vv3Err0Jia7F1Pe/Tzu0F6swGt1EIxMhUx5qFX8bVduHVMYwx15FZMb8r8QTtDvvElm4cjX1XPxSpOY0iRUiljxtB9QxNJ4uSG8kQOY4rh/7FQnkKmWjg14bAcMfofP2XACfocQxn+oaAq7fRnB3osJ5PopvfhdatsfnXqsAHZrD1Yg7AcxYyWsQND/WZhdgb38Cn+d6UUQDGMmh/FyQ+DVcbpWY2eXYdkBP8jJDeLlhv3vg0DI36qW6jhqi9xR30vGYEoT/nFvCPhNpYBxa/73g3N62/aMMUcd57T24lz5cwBks0mGhnLH3Meybaxku//9d9oyZ3Ds7LGMMabZRTTDyEgez2vM0j/5wFMs7kjwK++5qCHnPx+d6B/LfKd1Lyxa98Kidc9ttm3R1nZ+vNsnR2vUc7C5/L2bL9V48NEdPPPyIBf0tnDbT11IR/rYF9amXqG+98fUD2zHPfwKJj8CgBXP4PRcjBUIUdv1Q6iWsJJZ7FgaQlFMOYc3tBcrmsLpuYj67iex4q1Erv9lnEVrmH4BeIIXYOej8/Xv2yvnqHz/q9QPvkhw5UaCF/03nLbFp7yfqVUgEDzl38HprvtIxwpuDSuWOTbU8VzM5BDu+EHM5CAEwljRFuxoyu+miKamu1+OcEcOUNvxfWo7fwieR2DxxQSWXIqdXoSXH/HDHreGncz6gUYo6ne6TAzgTQxg8iN4hVFMKeevNRgBJ4iZCooArJYuwuvfTWDVWzG5YarbH6X2yhPgVnE6VxFcdx3u4V3UXv4eVqyFwAVX+kFOrfy6Dp1RqBZP+TU6pWCUQM86jFvDG9qHKZ/i627ZWKkOwLy2let157LTXXgjB47bqeN0ryNy7YewUx2YSoHyj/6e+itP+J8Mx3Gyy6FexR3eD28cHm3ZfphjB7AiSZz2JdjtS7EiST8UGzuIyY0cXWc0OTVLKIpXGPNDmfwYdqodu3UJTmsvwGshVzCKnWj1u53irVjxVqxoEgx++JYfAa+OFWvxf+44Ib9Tp5Tz7z8V1GEMdrLND3yObDFzq35nUyg2/f2WzSYZPDSMO7gHKxz3tyXa87935nz9uXYmTvX8SwFUA9z1lR+xdnGaD9+0riHnPx/Nh38sZ0PrXli07oVF657bFECduX379vG+972Phx56iN7e3mM+77ouf/Znf8a3v/1tcrkcK1as4M477+TKK688o8dRAHV8xhiefGGAv/s/O/AMbHrHKn7ikkUnfUfdy4+QSQUZd5OvdTvVKtR2P4n76jZMtegPpgWCa64luOYarEAI9/BOSv9xvx88vJ4TwApG/fkjwai/zSUY9bfguDVwaxi37r9YdGv+8fEMdiwDoQhUiv7WHcDOLsPpWInd0ok3OYQ33o8pjGFnevztKYl2qBRwR17FGzuIFYphpbJ+p4YxmGoBU/Frt5ygH1g4U9tLAkGyna0Mj1WwbBtjjD/LpVLEFMfwRg/ijh3EVAoEFq3B6b0YO/5aN4AxBlPO+YN5C6NYsTR2S5e/XegEX29Tr/oBytQvkx/By41gOUGcnnU4PRfiDe+n/B8PYMp5AkvWUz+wHdwqdnYFdqbHf+EdbfHXVsphihN4uUHMhL+tBieE3XJk0HAKKxSDcGxq3QGwHRIRm9zwsP91ft1WIjO1Hcn/NeGHJUde5jkB7FQnVqLVD2mKE/58mlPNlHFC/paeqfDKlCbBDhBYdhlWKEJ9/3P+eU5HMIKdaMNKtPrhVr3md/hNzcGxWzqwQnFqOx7HG9mPFUn6AZrtEFz1VjqvfS8TVuv06dzBPZSf+Du8kVf9ICsYxgon/JAkPhWUHPlzOO4HKaXJqa9bHbwaxnP9Li47MNXNY8B4YDzstqU4XRdMhx7GmKntZeOYasnvCDLe1NayMFYk4f+9TW1/8vKjuP0v4RXGCCxag92xAst2/K1bA7vwRl71/51FktixNHZ2+THfe+7wflo7Mkf/+zYeZmLA30objmNFEvNyXtRc/3l+tubDuk/1/Gv+x4hNUKu7ugqeiIiIzBlbtmzhd3/3d5mYOPGLyT//8z9ny5Yt/O3f/i3d3d089NBD3H777fzrv/4r3d3ds1jt/GRZFldd3MXqxWn+8l9e5K+/8zLff66f6zf08Oa1HYSOM1/UTrQRaktive4FixUME1p7Hay97oSP5XStIv6z91B7+Xv+C+kj6lX/xXWthKmWoVbyu1rAD5ucoL/VxA5gOQH/CkyFMeqDezC1ij+0NxzHeHXq+358vFVypNuKUNSfYXKWCtOndPxzvnFwsRPECoSpv/ID/7BUh397reLPRqkfZw5MKOa/mDeeH4JNhREYA7XyG5ZiYcUymFqJ2sv/MX2zne4m+q47cNqXTl3p6nFqe3+M27edenHitfUHI374kOrAWXEFVqLd71SbGMAbP4w5vNMPD98QEk33vhyp9fXrjaawk+1YHSte62Jygn730cRhvMKYH5RkevzQI92Fne7Gbun0A7Yj4dVUkOWVJqFe87dYGQ+7dTHBC66aniVjjOd3BhUnpgKfNnACfuA4OQjVon/lrvSik4Z7rxe86B24fc9Te/l72OlFBC96B3YsTSibhNd9nzsdK4j/zO+f8nwzxbIsrHgG4qe3rclOtGKvvvrY8wRCBHouhJ4LT3kOp33psf++LRsrvej0Cxc5zyiAaoBqzVMAJSIiInPCfffdxyOPPMIdd9zBJz/5yRMel8/n+djHPsbixf52oltvvZU//uM/5oUXXlAANYPaWiL85s9v4HtbD/K/n+njL//lJf7nd3dy7aXd3PjmJaTiM9PtYAUjhN5044yc63hMpYA7uAdvchA71eFvoYm1+N1JA7vwRg9gJbP+XJ3WXqiW8Y7MZrEdvyMqFAWsqc6rmh+QTXVixSM2+cm8H5AAVjgG4bi/BSnTg5XMgmXhjR6gfuB5vKE9fogWCE9349ipDr8rqDCON3EIb2LA746xbL/rx7L8P1u2vw0o2Y6VaPM7eeLpqZkyLt7wPup9LwAQuuRd08GQFUkQuuQnCV3yk/7XxK1hynk/qDvFsHjwu25wq363kFcHz6W9s5WRnDej85f8r1/8tMOV6ftY9nEHZTtti09r2+Hxz2kRWPwmAovfdFb3F5HzmwKoBlAHlIiIiMwVt9xyC7fffjsHDx486XF33XXXUR8/9dRT5HI51q07s5EDjdwamc0mG3bu2fa+G1PcesNatu8e5l9/uI9Hn3qV7/74IO+6aim3XH8BbS2vzYg6P9edhN7jXKmsMwPrLj7BfVad0SOkT+egjotgbYPnsnZeChddepoHt576kFPoiJ3zKeak8/P7vPG07oVlvq9bAdQM8zxD3TUEHQVQIiIicv7r7Ow84/ts376dX//1X+dXf/VXjzsv6mQ0A+rMLGqJ8P/85FrevXEx//Kj/Tzy+F4efnwPFy5r5a0XdXHDW5eTmzz7rWxz1Xz9+z4VrXth0boXlvmw7lPNgJrVlGR4eJjbb7+dyy+/nKuvvprNmzcfcKIOUAAAIABJREFU9zjXdfnTP/1Trr/+ei6//HJuvfVWnnzyyenPv/zyy6xdu5YNGzZM/3r/+98/W8s4qZrr7z9XB5SIiIjMR9/85jf5xV/8RW677TY++tGPNrucBWNRW5zbfupC/uhXruSmq5ZyeKTI/Y+8yAf+x7/xwCMv8uK+0YZdYEdERGQmzGoH1B133EFvby9PPPEEBw4c4CMf+QjLli3jp37qp4467lRDLrdv38769ev5+te/Ppvln5Za3Q+gAgqgREREZB4xxvDZz36Whx56iC984Qtcd92Jh1xL43Sko9xy7Uree80Kdh4Y5792jfCD5w7yw+cPk4qHWL04zZrFaVYvTtOTjWOfxuBnERGR2TBrAdT+/ft5+umn+fznP08kEmHVqlV88IMf5Gtf+9oxAdSphlw+//zzXHzxifaON9eRAEpb8ERERGQ+2bx5M4888gj/+I//yPLly5tdzoJnWxZrlmT4icuX8LPXLOfZXcM8u3OYHQfGeeblQQBi4QCrF6dZ2pWkVvcolGvU6x6Xr+ngkgvaFE6JiMismrUAaufOnaTTadrb26dvW7FiBTt27Djm2FMNuXz++eeJRqPceOON5HI53vzmN/Pbv/3bdHUdZ9DhLNMWPBEREZkP+vv7uemmm7j//vu55JJLuP/++3Fdl1tuueWo4z71qU/xnve8p0lVCkAo6LBxXScb13VijGFkoswrfeO8cmCcHQcmeHbXMI5tEY8G8TzDE88fpiMT5R2X97J+ZRvZdBRLYZSIiDTYrAVQhUKBaDR61G3RaJRyuXzS+x1vyGVbWxvr16/nAx/4AK7rcu+99/KRj3yEf/7nf8ZxTu+SpI26AkvJ9ffet2Xi836C/RsttPUeoXUvLFr3wqJ1y0LS29t71BuD3d3dbN26dfrjbdu2NaMsOUOWZdGejtKejvLWixcBfod+wLGwLIu66/Ffrwzxf/7zAP/w7zv5h3/fSSIaZEV3iouWtXLpqnay6egpHkVEROTMzVoAFYvFjgmbSqUSicSJg6BvfvObfOpTn+JjH/sYt9122/Ttf/EXf3HUcb/7u7/LVVddxe7du1m9evVp1dOoK7AMDuUBKBUrc36C/ZmYDxP7z4bWvbBo3QuL1j23neoqLCILyes78wOOPd0t1TeYZ9fBCfb0T7Lr4ATbdu/kH767k95snGWLUmRbIrS3REnGg0SCASIhh7aWCNGwLqQtIiJnbtb+91i1ahVjY2OMjo7S2toKwO7du7nggguOOfZkQy5HRkZ44IEH+JVf+RXS6TQA1WoVgEgkMgsrOTltwRMRERGRuaC3I0FvR4LrN/QAMDBW5Lmdwzy3e4Tte0aYyFePuU8wYLNhVTtXXdTFRctbCWjuqYiInKZZC6CWLVvGhg0b+NznPsfv//7v09/fz4MPPnjcy/eebMhlS0sLjz76KIVCgd/5nd+hVCrxqU99iquuuoolS5bM1nJOaHoIuQIoEREREZlDOjMxbti4hBs2+s+pqzWXkckyhVKdcrVOqery8qtjPP3iAE+/NIhjW2SSYVpTEToyUVZ0p1ixKEVPNo5j67mwiIgcbVb7Z7/4xS9yzz338La3vY1gMMimTZu49dZbz3jI5f33389nPvMZrr32WowxXH/99XzmM5+ZzaWc0JEAKqAASkRERETmsFDQYVFb/Kjb3ry2g59/xyq27xlhT/8kI5NlRifKPLtzmB9sOwT42/yy6QidmRhdrTFWL0mzZnFaW/dERBa4Wf1foKOjg82bNx9z+5kOuVy5ciUPPPDAjNc3E6Y7oNSOLCIiIiLzUMCx2bAqy4ZV2enbjDEMjZfY3T/JqwM5BsdKDI2XeH7vKP/29KvYlsXijgSRkINtWziORSYRpn1qzlQ0EiAcsAkFHbrb4wqrRETmIf1kn2E11wW0BU9EREREFg7LsujIxOjIxLjqoq7p22t1l10HJ3lp/xh7D01Sr3vUXI9SxePAQJ6JwrFzpmzLYkVPiouXtbJsUZJsOkp7S1TPr0VE5jgFUDNMM6BERERERHzBgMO6pRnWLc0c9/OVmsvoZJly1aVacylVXHb3T/DC3lG+9YO9HLlmtQXEIgECAZugY9OWjtKZjtCTTbCsK8nyRSkNRBcROc8pgJph9ekAymlyJSIiIiIi57fwceZMXbqqnZ+9biWFco1Dw0UGx4sMjpUolOvU6i7Vukeh7PLjHUN8/7lD0+dZtbiFld0tJKJB4pEA0XCAgGPj2BbhkENvNn7Uc/S66zEyUSabjmLb1qyuW0RkIVIANcNemwGl/8RERERERM5WPBLkgt4WLuhtOeZz2WySwcFJxvNV9vRP8NL+MV7aP8bze0ZPeD7HtljSmaQnG+fwSJH9AzlqdY+OTJQbNy7h6ou7CAX1JrKISKMogJphNVdb8EREREREGs2yLDLJMJev6eDyNR2A39VUKNcplmsUK3Vc1+BO3bb30CS7D07w7M5hutpivG1DD9l0lB8+f4i/fXQH33x8D29a0cbyRSmWL0rh2BYThQqThRrpZIhVPWnCIQVUIiJnSwHUDDvSAaU96CIiIiIisyvg2LTEQ7TEQ8d87oq1Hce9z9sv6+GVA+N898d9PL93lB8+f/i4xzm2xcruFB2tMYxn8IwhGLBpiYdJJ0K0piL0ZhO0psLUXcPWnUN879l+9h/O8daLu7hx4xLaWiIzul4RkblEAdQMq7kewYCNZWkLnoiIiIjI+c6yLNYsybBmSQZjDKOTFfYdnsSyLFLxEKlYkMHxEi/t87f5vbB3FNvy71ere0wWqxjz2vmi4QC2BYVynbZUhHXLMmzZepAtWw+ycV0HFy5rZfmiFF1tMWy9ZhCRBUQB1Ayr1T1C2n4nIiIiIjLnWJZFW0vkmE6ljkyMi5e3Hfc+rucxWagxPFGib6hA32CeSs3lyos6uXBZK7ZlMTJR5tGnX+UH2w/xoxcGAAiHHFqTYVriIdLJMD3tcZZ2JlnSmSR1nA4uEZG5TgHUDKvXPYIaXigiIiIisiA4tk0mGSaTDLOqN33cY9paImx652r+73es4tBIgT2HJnl1IM94rsJEocrOA+M8ORVMAWSSYZZ2JlnalaSzNUprMkJrKkymNX7c84uIzAUKoGaYOqBEREREROR4bNuiJ5ugJ5s45nPFco1XB/LsH8j5vw7neG7XMK/b3YdlQUvcnzfVEg+RjIVIxoKkE2G62mIsao2RiocoV12K5RrVukfQsQkGbKLhANGwXv6JSPPoJ9AM82dAqQNKREREREROXywSZO3SDGuXZqZvq1RdhifLjE2WGc1VKNcNfYcnGc2VGRwvsad/knyphuuZk5z5NUs6E7xpRRtrlqTxPD/0KlddFrXFWLYoRVg7OUSkgRRAzbBa3SMUVAeUiIiIiIicm3DIoac9Tk+7v/Uum00yNJQ76hhjDBOFKodHihwaLZIrVomGA8TCAUJBh3rdo+Z6jOcrvLhvjH976lX+5Uf7j3ksx7bozSZoSYQIBvyuqcUdCS5Z0UZ3e1wXWRKRc6YAaob5W/D0zoGIiIiIiDSeZVmkE2HSifBR3VPH856rl1Ms19l/eJJQ0CEWCRAKOBwYyrP74AR7D00yUahSr3uUqy5PvjDA/9qym9aUP5MqnfCHpoeCDq7n4XmGgGOTiAVJxkK0pSJ0t8dwbL0hLyLHUgA1w2p1j2BIAZSIiIiIiJx/YpEA65a1HnVbW0uESy9oP+bY0cky2/eMsH3PKAOjRXb2TZAv1U56/lDQZllnkiVdSbItUdpaIiRjQQqlOpPFKsVyHcvy52EFAzYX9LTQow4rkQVBAdQMq7keMSfY7DJERERERETOSWsqwnWX9nDdpT3Tt9Vdj1rdw7EtHMeiVvfIFWvkijUGx4rs6Z9kz6FJvv9sP9W6d5qPE+bi5W10tcZIxoIkY0GK5TqjuQqjk2XaW+O0RAN0t8XpbI0SCellrMhcpH+5M8yfAaUOKBERERERmX8Cjk3AeW2LnROyiYQCZNNRVnSnuPKiLsCfTZUr1RiZKJMv1UhE/WApHvHfrHc9Q7FS5+X9Y2zbPcLTLw1QrrrHPF40HKBSc/FeN2g9HgnQ1hIhkwgTiwSJRwMkIkHi0SCxSIBULERHJkprKqztgCLnEQVQM6xW9wgG9ENOREREREQWLsuySMVCpGKhEx6TiAbpSEe5dn03xhjKVZfJYpVcsUYsHKA1FSYSCpDOxHnhlQH6R4oMjZcYnigzPFFiLF+hb6hAYepqfm/k2BYdmSgre1pY1dvC8q6U/1rN8uuzAAt/O2A6GcbWNkCRhlIANcPUASUiIiIiInJmLMsiGg4QDQfofMMs9WDApieboCebOOH9665HsVKnWK4znqswOF5iaLxE32Cera8M8YNth076+NGww7KuFMsXpYiGX3s9F48EaU2FySQjREKOP7/KskhEg3rdJ3KGFEDNsLqrDigREREREZHZFHDs6Y6rrtbYUVcE9Izh0EiRvsE8nmcwGIzB/4WhXvc4MFRgT/8E//bUq3jGnOSRfLZlsag9xtLOJJGQw8BokcOjJQA2rGrnirUdXNDTgm2ffleV63naMijzmgKoGaYOKBERERERkfOHbVn0tMfpaY+f8ti66+F5hiO78XLFGqO5CmO5CpWqOx1ejUyU2T+Q44W9o1TrHl2tUVYtbqFSdfmPZ/v59x/3Tc/KOjK/KhENkIiFSEQCOI7NkWxqolBlLFehUK7TmYmybmmGtUszXGbZ2MZoa6DMGwqgZljN9QipA0pERERERGTOCTg2vK6foDXl0JqKnNE5SpU62/eMsPfQJLZt4dgWxkC+5F8tsFCqUXM9jGcwQDYdZdXiNIlIkFcHcjz54gD/8Ww/fOsFQgGbrrYYyWgQ27ZxbItYJEB7S4S2VIRQ0KFUrVOq1KnVPI70bgUci5Z4mHQiRDoRpiURIhENYlkWxhiqdY9y1SUeCRw1VF6kkRRAzSBjzNQQcnVAiYiIiIiILETRcICN6zrZuK7zrO7veh6vDuQZL9XZsXeE/uECpWodz6vjuob9AzXGcxVOvVHwaI5tEY8EKFZc6q43fXssHKAlEWJldwvrlmVYtzRDOhE+q9pFTkYB1Ayqu/6PgFBQCbKIiIiIiIicOce2Wb4oRTabZMOK1uMeU3c9RnMVajV3enh7OOj4V/gDqnWPiUKV8VzF/z1fYTxfoViuEwsHiEeDhIMOxXKNiUKV0ckKW3cO8YPt/rD2dCJEd3uc7rY4qXiIcMghEnT830MO4ak/h4MOkVDAvy3kYE91WOVLNYbGy5SqdTrSUdpSkTOahyXzkwKoGVSr+ymyOqBERERERESkUQKOTUc6esLPh4MOHenoSY95I88zHBjM89L+MQ4O5Tk4XODxbYeo1NzTur8FRMIBPGOoVI++T8CxaGuJEo8EiIYcopEg2XSEzkyMzkyUZMzfIhiLKKKYz/S3O4Nq7pEASh1QIiIiIiIiMnfYtsXSriRLu5LTtxljqLuGSs2lUnUpT/1eqdb9P099XKq4lCr+LCosyLZEaU9HiIQCDI2XODxaZHiiTLlSp1StMzxRZusrQ7jesRsJ08kwmUSI1mRkaneR3zlVLNcYL1SZyFcIhwJ0t8Xobo8TjwSp1FzK1TrG+OFbJOwQdGw8z+Aag2NZtKYiZNNRsumogq4m0Vd9BtXqfsqrIeQiIiIiIiIy11mWRTBgEQzYJKLBszrHuqWZ497ueh4jkxWGxkrkSlUKpTq5YpVy3XBwMEf/SGF6l5ExEIsESCfCLM4mKFXr9A8X2LZ7ZDrECgVtLMs6pvvqeLLpCMu6Uixqi5Er1hgcLzGWq5CKBadDqkQsSDQUIBp2aGuJ0pmJamD7OVIANYOmt+AFtQVPRERERERE5EQc2z7uNsFsNsnQUO60zlF3PWp1j3DQmZ4x5RlDteZSq3s4to1t+/OaRybKDI2XGBgrsv9wjr2HJvnPlweJRwJk037ANFms8tzuESYL1ePUa9GRiZJJhomFA8QiAYIBB8e2sC2LSMghnQyTTvhXH8wkw9NXHhSfAqgZdCSAUgeUiIiIiIiISGMFHPuYriQ/DAoQCR19bCIaPGp7ITB1FftjX79Xqi7FSp1ytU6xXGdwvET/cIH+4QKTxSpjuQqlSp1a3cP1DJ5nqNa9Y84TcCzi0SAYfzujN/W7Mf6Wx2w6QmdrjI50lFgszESuhOsaWuIh2tNR2lsitLVESM6TIEsB1Aw6MgMqpA4oERERERERkfPaieY3h6eu6gdhAFb2tJzyXHXXm7raoH/1wbF8hfFchUK5hmVZU7/Axv+97noMjpd45cA4T74wAPiBlW1bVGtHh1mhoE17iz+7KjgVunmeR7nqUq66fhZh/O6vgGMTjwaIR4Kk4iG/yywTJRoO0D9coG8wT831+MhPXzTrVyZUADWDwlNXv2tJhE5xpIiIiIiIiIjMFwHHD4naW07/yoNH1F2Pzo4UIyN5AIrlOiOTZYYnSgxPlBmZKDM8UaZUqVN3PUqVOo49te0vEcaZCq4soOYaiuUaE/kq+w/nmHjDdsKWeIg1S9J4xmCjAGrO6u1IcO9tb2HV4sxp71kVERERERERkYUr4NhHdSPFIgFikQSLOxLnfO5ytc7QeJliucai9jipWPMaZhRAzbDu9nizSxARERERERERIRIKzEiQNRM0LVtERERERERERBpKAZSIiIiIiIiIiDSUAigREREREREREWkoBVAiIiIiwr59+9i4cSN9fX2nPPbTn/40d9999yxUJSIiIvOFAigRERGRBW7Lli1s2rSJiYmJkx43Pj7Ob/3Wb/F3f/d3s1SZiIiIzBcKoEREREQWsPvuu48/+ZM/4Y477jjpceVymXe9610kk0luvPHGWapORERE5otAswsQERERkea55ZZbuP322zl48OBJjwsGgzz88MNks9lz2n7X1ta4S0Fns8mGnft8pnUvLFr3wqJ1Lyzzfd0KoEREREQWsM7OztM6znEcstnsOT/eyEgezzPnfJ43ymaTDA3lZvy85zute2HRuhcWrXthmQ/rtm3rpG80aQueiIiIiIiIiIg0lAIoERERERERERFpKAVQIiIiIiIiIiLSUAqgREREROQY/f39bNiwgWeeeabZpYiIiMg8oCHkIiIiIkJvby87duyY/ri7u5utW7ce99jPfvazs1WWiIiIzBPqgBIRERERERERkYZSACUiIiIiIiIiIg2lAEpERERERERERBpqwc6Asm1rTp//fLQQ1wxa90KjdS8sWvfcNR/WMF818u9mof69a90Li9a9sGjdC8tcX/ep6reMMWaWahERERERERERkQVIW/BERERERERERKShFECJiIiIiIiIiEhDKYASEREREREREZGGUgAlIiIiIiIiIiINpQBKREREREREREQaSgGUiIiIiIiIiIg0lAIoERERERERERFpKAVQIiIiIiIiIiLSUAqgRERERERERESkoRRAiYiIiIiIiIhIQymAmiHDw8PcfvvtXH755Vx99dVs3ry52SU1xLZt2/iFX/gFrrjiCq655hruvfdeSqUSAFu2bOHd734369ev5+abb+aZZ55pcrUzz3VdPvCBD3D33XdP3/aNb3yDd7zjHVx66aVs2rSJXbt2NbHCmTU+Ps5v//Zvc+211/KWt7yFu+66i0KhAMzvv+8XXniB97///VxxxRX8xE/8BH/4h39ItVoF5ue69+3bx8aNG+nr65u+7WTrrNfr/NEf/RFXXnkll112GXfeeSf5fL4ZpZ+T4637q1/9Ku985zu57LLL+Omf/mm+853vTH9uPq/7iFdeeYX169fz1FNPTd9WKBS466672LhxIxs3buSee+6hVqvNZskiJ6XnYPPz/6Y30nMwPQebj+vWczA9BztiwTwHMzIjfuEXfsHcfffdplQqmVdeecVcf/315uGHH252WTNqcnLSbNy40fz1X/+1qdVq5tChQ+aWW24x99xzj9m7d6+55JJLzJYtW0y1WjV///d/bzZu3GgKhUKzy55RX/jCF8zatWvNJz7xCWOMMU8++aS5/PLLzbPPPmsqlYr50z/9U/Oud73LeJ7X5Epnxvve9z7zG7/xGyaXy5nR0VGzadMm8wd/8Afz+u/b8zxzzTXXmL/5m78xruuaQ4cOmRtvvNF86Utfmpfrfuyxx8xVV11lVq9ebQ4cOGCMMadc5xe/+EXz3ve+1wwMDJixsTHz4Q9/2Pze7/1eM5dxxo637m9+85vm6quvNi+99JLxPM889thj5k1vepN57rnnjDHzd91HFItFc9NNN5nVq1ebJ598cvr2T3ziE+bDH/6wmZycNP39/ebmm282X/n/27v7mCrLP47jHx4OApOBPQjVH5GHI1AQLI80YxkeNpcBwxzmjJFIc2HOB7BSU9vM6dQFOB0aUWs9sMpwYpIz2Vqtmpv5UNkfKqAUOnwYESMMFLl+f/TjjCOkv19xOHD7fm1unOu6D36/54Zzf3adm/t+883hLh34W2Qw6x2bBkMGI4NZrW8yGBmsz+2UwTgDagj88ssvOnz4sJYvX67g4GA5HA7NmzdPVVVVvi5tSJ0/f15Op1P5+fkKDAxUVFSUsrOz9f3332vPnj2aPHmy0tLSZLPZNHfuXEVGRurzzz/3ddlD5tChQzpw4ICefPJJ99inn36qjIwMJSUlKSgoSEuWLFFra6sOHTrkw0qHxk8//aSTJ09qw4YNGjt2rMaNG6etW7cqPz/f0vv7jz/+0MWLF9Xb2ytjjCTJ399fwcHBlut7586deuONN1RUVOQxfqs+q6urtWDBAo0fP14RERFavny59uzZ4/5kdqT7u75bW1u1cOFCxcXFyc/PT9OmTZPdbtexY8ckWbfvPuvWrVN6eroCAgLcY11dXaqtrdWSJUsUFhame+65R4sWLbLc8Q2jFxmMDEYGs87+JoORwchg1s9gLEANgfr6ekVEROiuu+5yj02YMEGnTp3yYVVDLy4uTuXl5e7HxhjV1dUpISFBDQ0NcjgcHtvb7XbLvAatra1avXq1SkpKFBoa6h6/se+AgABFR0dbou8TJ07Ibrfrww8/VHp6uh5//HFVVFQoMjLS0vs7LCxMBQUF2rx5sxITE/XEE0/ogQceUH5+vuX6njVrlmprazVlyhSP8Zv12dHRoQsXLigmJsZj7urVq2pqahqOsv+1v+u7oKBAubm57sdnzpxRfX29EhISLN23JNXU1KipqUmLFy/2GG9qatK1a9c8+p4wYYIuXryo33//3es1A7dCBiOD9SGDjf6+yWBksD5kMOtmMBaghkBnZ6dCQkI8xkJCQtTV1eWjiryvp6dHa9euVXNzs5YuXarOzk4FBwd7bBMcHOy+NsFo1tvbq5dffln5+fl68MEHPeas3Hd7e7tOnjypS5cuad++ffrkk090/Phxbd682dJ99/b2KjAwUGvWrNEPP/yg2tpaNTQ0aNu2bZbrOzIyUn5+fgPGb9Zn3ydN/d/zxowZIz8/v1HzOvxd3/01NzfrhRdeUHZ2tpxOp6X7bmxsVFlZmUpKShQYGOgx13d9hf59931t5WMcRg8yGBmsP6v0TQYjg/VHBrNu37djBmMBagiEhoYO+CH4888/NXbsWB9V5F2tra16/vnn9eOPP6qqqkqRkZGDhr2uri5LvAYVFRUKCgrSc889N2AuJCRE3d3dHmNW6XvMmDHy9/fXihUrFBoaqnvvvVcLFy7UgQMHLL2/6+rqVFdXp9zcXAUFBcnhcGjRokX66KOPLN13fzfrc7ADX3d3t4wxlnkdvvnmG+Xk5CgtLU3r16+XNPgB3wp9d3d3q6ioSKtWrdJ99903YL7vbIP+ffeFvdHcN6yDDEYG688qfZPByGD9kcGs2fftmsFYgBoCDodDbW1t+u2339xjjY2NHqfLWcWpU6c0a9YshYeH6+OPP3b/skycOFFnz5712LahocESr8HevXt1+PBhOZ1OOZ1O7d27V7W1tcrKypLD4dCZM2fc216/fl1NTU0DTp0djWJiYmSMcd95RPrrU1fJ2vu7paXF3WefwMBA2Ww2S/fd3836DA8P1/jx4z1+7hsaGmSz2RQdHT3MlQ69d999V4sXL9Yrr7yi1atXy9//r8OkVfs+ceKEmpqatGbNGvd73PXr11VYWKi33npL0dHRstlsHn03NjYqKipqVIcfWAcZzNrHZDIYGYwMRgazat+3bQbz1dXPrWbOnDlmxYoVprOz09TX15u0tDSza9cuX5c1pC5fvmwee+wxs3HjxgFz9fX15uGHHzZ1dXXuOzZMmjTJtLW1+aBS73r11Vfdd2D59ttvjdPpNEeOHHHfgSU9Pd1cvXrVx1X+e93d3SY9Pd0UFxebK1eumAsXLpiZM2eaDRs2WHp/19fXm4SEBFNRUWF6enrMr7/+ajIzM82WLVss23dzc7PHHTlu1WdpaanJzs42LS0t7juRvPTSS75s4R+5se/du3ebpKQkc+zYsUG3t2rfN4qPj/e4A0txcbEpKCgwbW1tpqWlxWRnZ5uysrLhKhe4JTKYNY9NgyGDkcGs1jcZjAzW3+2QwViAGiIXL140ixYtMikpKSY1NdWUl5f7uqQht337djNx4kSTlJRkkpOT3f+eeuopY4wxX331lcnMzDTJycnm6aefNkeOHPFxxd7RP/wYY0x1dbWZPn26SU5ONrm5uaahocGH1Q2tlpYWs3jxYjNlyhSTkpJi1q1bZ7q6uowx1t7f3333nZk9e7aZNGmSmTZtmikrK3MHWiv2PdjB8GZ9dnd3m02bNpnU1FTjdDpNcXGx6ejo8EXp/8qNfbtcLhMfH+/x/pacnGx27txpjLFu3ze6Mfx0dHSYlStXut8H1q9fb65duzZc5QK3RAaz5rFpMGQwMpjV+iaDkcH6ux0ymJ8x/73HJQAAAAAAAOAFXAMKAAAAAAAAXsUCFAAAAABuONLKAAAEIklEQVQAALyKBSgAAAAAAAB4FQtQAAAAAAAA8CoWoAAAAAAAAOBVLEABAAAAAADAqwJ9XQAADMblcun8+fODzjkcDtXW1nq9htjYWG3ZskXZ2dle/78AAABGAjIYAG9hAQrAiLVgwQLNmzdvwHhgIG9dAAAA3kIGA+ANvIMAGLFCQ0N19913+7oMAACA2woZDIA3cA0oAKPSuXPnFBsbq3379mnGjBlKSkpSXl6eTp065d6mp6dHlZWVmj59uhITE5WVlaX9+/d7fJ+vv/5as2fPVlJSklwul95++22P+cbGRuXl5SkxMVEul0vV1dXD0h8AAMBIRAYD8E+xAAVgVNu0aZOWLVum6upqhYWFaf78+ero6HDPvfPOOyouLtZnn32mjIwMFRcX64svvpAkHT9+XIWFhUpNTVVNTY1WrVql8vJy7dq1y/39q6qqNHfuXO3fv18ul0tr165Vc3OzT3oFAAAYKchgAP5ffsYY4+siAOBGLpdLly5dks1mGzC3cuVKpaamKj09XWvWrFFeXp4kqaOjQ1OnTtWKFSuUmZmpRx99VK+99prmzJnjfu6yZcvU3Nys3bt3q7i4WJcvX9YHH3zgnq+pqVFAQICysrIUGxurwsJCFRUVSZLa29uVkpKi7du3a/r06V5+BQAAAIYfGQyAt3ANKAAjVm5urp599tkB43fccYfa29slSZMnT3aPh4WFyW636/Tp0zpz5ox6enr0yCOPeDx38uTJ+vLLLyVJp0+f1tSpUz3mZ86c6fE4Ojra/XV4eLgkqaur6583BQAAMMKRwQB4AwtQAEas8PBw3X///YPO9YWfGz+d6+3tlb+/v4KCggZ93vXr1913cPlf7uTi7z/wL5U5cRQAAFgZGQyAN3ANKACj2s8//+z+ur29XWfPnlV8fLyio6Nls9l09OhRj+2PHj2qmJgYSZLdbvd4viSVlZXpxRdf9H7hAAAAoxgZDMD/izOgAIxYV65c0eXLlwed6/sErLS0VHfeeafGjx+vkpISjRs3TjNmzFBwcLDmz5+vrVu3KiIiQnFxcTp48KAOHjyo0tJSSVJBQYFycnK0Y8cOZWRk6OTJk3r//fe1evXqYesRAABgpCGDAfAGFqAAjFiVlZWqrKwcdK7vVrzPPPOMXn/9dV26dEkpKSl67733FBoaKklaunSp/P39tXHjRrW1tclut6u0tFQzZsyQJD300EPavn27tm3bph07digqKkpFRUXKyckZngYBAABGIDIYAG/gLngARqVz584pPT1dVVVVcjqdvi4HAADgtkAGA/BPcQ0oAAAAAAAAeBULUAAAAAAAAPAq/gQPAAAAAAAAXsUZUAAAAAAAAPAqFqAAAAAAAADgVSxAAQAAAAAAwKtYgAIAAAAAAIBXsQAFAAAAAAAAr/oPYvr3QQpsVC4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1440x576 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2) = plt.subplots(nrows = 1, ncols = 2, figsize = (20,8))\n",
    "\n",
    "ax_1=sns.lineplot(history.epoch,history.history['accuracy'],ax=ax1,label='Train')\n",
    "sns.lineplot(history.epoch,history.history['val_accuracy'],ax=ax1,label='Test')\n",
    "ax_1.set(xlabel='Epoch', ylabel='Accuracy')\n",
    "\n",
    "ax_2=sns.lineplot(history.epoch,history.history['loss'],ax=ax2,label='Train')\n",
    "sns.lineplot(history.epoch,history.history['val_loss'],ax=ax2,label='Test')\n",
    "ax_2.set(xlabel='Epoch', ylabel='Loss')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Config. 2: reducing the batch size and learning rate in order to gain more stability but updating the weights more frequently"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:19:59.341661Z",
     "start_time": "2020-03-01T09:19:40.017881Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Falling back from v2 loop because of error: Failed to find data adapter that can handle input: <class 'pandas.core.frame.DataFrame'>, <class 'NoneType'>\n",
      "Train on 6844 samples, validate on 1711 samples\n",
      "Epoch 1/50\n",
      "6844/6844 [==============================] - 2s 219us/sample - loss: 1.5471 - accuracy: 0.3824 - val_loss: 1.3182 - val_accuracy: 0.4833\n",
      "Epoch 2/50\n",
      "6844/6844 [==============================] - 1s 123us/sample - loss: 1.2845 - accuracy: 0.5022 - val_loss: 1.2468 - val_accuracy: 0.5278\n",
      "Epoch 3/50\n",
      "6844/6844 [==============================] - 1s 123us/sample - loss: 1.2372 - accuracy: 0.5291 - val_loss: 1.2279 - val_accuracy: 0.5395\n",
      "Epoch 4/50\n",
      "6844/6844 [==============================] - 2s 220us/sample - loss: 1.2171 - accuracy: 0.5298 - val_loss: 1.2166 - val_accuracy: 0.5319\n",
      "Epoch 5/50\n",
      "6844/6844 [==============================] - 1s 123us/sample - loss: 1.2052 - accuracy: 0.5323 - val_loss: 1.2079 - val_accuracy: 0.5377\n",
      "Epoch 6/50\n",
      "6844/6844 [==============================] - 1s 179us/sample - loss: 1.1997 - accuracy: 0.5361 - val_loss: 1.2083 - val_accuracy: 0.5324\n",
      "Epoch 7/50\n",
      "6844/6844 [==============================] - 2s 243us/sample - loss: 1.1935 - accuracy: 0.5377 - val_loss: 1.2145 - val_accuracy: 0.5336\n",
      "Epoch 8/50\n",
      "6844/6844 [==============================] - 1s 119us/sample - loss: 1.1890 - accuracy: 0.5434 - val_loss: 1.2049 - val_accuracy: 0.5383\n",
      "Epoch 9/50\n",
      "6844/6844 [==============================] - 2s 281us/sample - loss: 1.1847 - accuracy: 0.5416 - val_loss: 1.2041 - val_accuracy: 0.5324\n",
      "Epoch 10/50\n",
      "6844/6844 [==============================] - 1s 181us/sample - loss: 1.1795 - accuracy: 0.5419 - val_loss: 1.2049 - val_accuracy: 0.5307\n",
      "Epoch 11/50\n",
      "6844/6844 [==============================] - 1s 205us/sample - loss: 1.1776 - accuracy: 0.5406 - val_loss: 1.2035 - val_accuracy: 0.5348\n",
      "Epoch 12/50\n",
      "6844/6844 [==============================] - 2s 231us/sample - loss: 1.1755 - accuracy: 0.5414 - val_loss: 1.2044 - val_accuracy: 0.5283\n",
      "Epoch 13/50\n",
      "6844/6844 [==============================] - 1s 185us/sample - loss: 1.1732 - accuracy: 0.5408 - val_loss: 1.2037 - val_accuracy: 0.5260\n",
      "Epoch 14/50\n",
      "6844/6844 [==============================] - 1s 137us/sample - loss: 1.1680 - accuracy: 0.5470 - val_loss: 1.2023 - val_accuracy: 0.5213\n",
      "Epoch 15/50\n",
      "6844/6844 [==============================] - 2s 236us/sample - loss: 1.1687 - accuracy: 0.5406 - val_loss: 1.2031 - val_accuracy: 0.5231\n",
      "Epoch 16/50\n",
      "6844/6844 [==============================] - 1s 187us/sample - loss: 1.1660 - accuracy: 0.5412 - val_loss: 1.2019 - val_accuracy: 0.5237\n",
      "Epoch 17/50\n",
      "6844/6844 [==============================] - 1s 156us/sample - loss: 1.1639 - accuracy: 0.5446 - val_loss: 1.2043 - val_accuracy: 0.5266\n",
      "Epoch 18/50\n",
      "6844/6844 [==============================] - 2s 232us/sample - loss: 1.1600 - accuracy: 0.5473 - val_loss: 1.2071 - val_accuracy: 0.5219\n",
      "Epoch 19/50\n",
      "6844/6844 [==============================] - 2s 234us/sample - loss: 1.1587 - accuracy: 0.5437 - val_loss: 1.1993 - val_accuracy: 0.5278\n",
      "Epoch 20/50\n",
      "6844/6844 [==============================] - 1s 200us/sample - loss: 1.1582 - accuracy: 0.5438 - val_loss: 1.2069 - val_accuracy: 0.5237\n",
      "Epoch 21/50\n",
      "6844/6844 [==============================] - 2s 281us/sample - loss: 1.1546 - accuracy: 0.5492 - val_loss: 1.2058 - val_accuracy: 0.5213\n",
      "Epoch 22/50\n",
      "6844/6844 [==============================] - 1s 146us/sample - loss: 1.1536 - accuracy: 0.5498 - val_loss: 1.2037 - val_accuracy: 0.5219\n",
      "Epoch 23/50\n",
      "6844/6844 [==============================] - 2s 238us/sample - loss: 1.1513 - accuracy: 0.5526 - val_loss: 1.2070 - val_accuracy: 0.5225\n",
      "Epoch 24/50\n",
      "6844/6844 [==============================] - 1s 172us/sample - loss: 1.1491 - accuracy: 0.5507 - val_loss: 1.2098 - val_accuracy: 0.5213\n",
      "Epoch 25/50\n",
      "6844/6844 [==============================] - 2s 285us/sample - loss: 1.1502 - accuracy: 0.5456 - val_loss: 1.2044 - val_accuracy: 0.5237\n",
      "Epoch 26/50\n",
      "6844/6844 [==============================] - 1s 169us/sample - loss: 1.1477 - accuracy: 0.5519 - val_loss: 1.2108 - val_accuracy: 0.5167\n",
      "Epoch 27/50\n",
      "6844/6844 [==============================] - 2s 234us/sample - loss: 1.1468 - accuracy: 0.5494 - val_loss: 1.2113 - val_accuracy: 0.5172\n",
      "Epoch 28/50\n",
      "6844/6844 [==============================] - 1s 179us/sample - loss: 1.1441 - accuracy: 0.5539 - val_loss: 1.2081 - val_accuracy: 0.5184\n",
      "Epoch 29/50\n",
      "6844/6844 [==============================] - 1s 185us/sample - loss: 1.1397 - accuracy: 0.5544 - val_loss: 1.2248 - val_accuracy: 0.5073\n",
      "Epoch 30/50\n",
      "6844/6844 [==============================] - 1s 172us/sample - loss: 1.1418 - accuracy: 0.5535 - val_loss: 1.2098 - val_accuracy: 0.5213\n",
      "Epoch 31/50\n",
      "6844/6844 [==============================] - 2s 223us/sample - loss: 1.1394 - accuracy: 0.5501 - val_loss: 1.2062 - val_accuracy: 0.5213\n",
      "Epoch 32/50\n",
      "6844/6844 [==============================] - 1s 139us/sample - loss: 1.1384 - accuracy: 0.5530 - val_loss: 1.2125 - val_accuracy: 0.5202\n",
      "Epoch 33/50\n",
      "6844/6844 [==============================] - 1s 137us/sample - loss: 1.1372 - accuracy: 0.5560 - val_loss: 1.2090 - val_accuracy: 0.5172\n",
      "Epoch 34/50\n",
      "6844/6844 [==============================] - 1s 124us/sample - loss: 1.1351 - accuracy: 0.5527 - val_loss: 1.2171 - val_accuracy: 0.5207\n",
      "Epoch 35/50\n",
      "6844/6844 [==============================] - 1s 177us/sample - loss: 1.1354 - accuracy: 0.5525 - val_loss: 1.2172 - val_accuracy: 0.5161\n",
      "Epoch 36/50\n",
      "6844/6844 [==============================] - 2s 237us/sample - loss: 1.1342 - accuracy: 0.5548 - val_loss: 1.2224 - val_accuracy: 0.5120\n",
      "Epoch 37/50\n",
      "6844/6844 [==============================] - 1s 173us/sample - loss: 1.1315 - accuracy: 0.5533 - val_loss: 1.2181 - val_accuracy: 0.5248\n",
      "Epoch 38/50\n",
      "6844/6844 [==============================] - 2s 245us/sample - loss: 1.1313 - accuracy: 0.5545 - val_loss: 1.2145 - val_accuracy: 0.5149\n",
      "Epoch 39/50\n",
      "6844/6844 [==============================] - 1s 125us/sample - loss: 1.1305 - accuracy: 0.5545 - val_loss: 1.2240 - val_accuracy: 0.5114\n",
      "Epoch 40/50\n",
      "6844/6844 [==============================] - 1s 214us/sample - loss: 1.1282 - accuracy: 0.5520 - val_loss: 1.2189 - val_accuracy: 0.5260\n",
      "Epoch 41/50\n",
      "6844/6844 [==============================] - 1s 149us/sample - loss: 1.1255 - accuracy: 0.5586 - val_loss: 1.2159 - val_accuracy: 0.5155\n",
      "Epoch 42/50\n",
      "6844/6844 [==============================] - 2s 238us/sample - loss: 1.1280 - accuracy: 0.5570 - val_loss: 1.2152 - val_accuracy: 0.5207\n",
      "Epoch 43/50\n",
      "6844/6844 [==============================] - 1s 151us/sample - loss: 1.1248 - accuracy: 0.5608 - val_loss: 1.2179 - val_accuracy: 0.5172\n",
      "Epoch 44/50\n",
      "6844/6844 [==============================] - 2s 247us/sample - loss: 1.1233 - accuracy: 0.5598 - val_loss: 1.2186 - val_accuracy: 0.5202\n",
      "Epoch 45/50\n",
      "6844/6844 [==============================] - 2s 244us/sample - loss: 1.1227 - accuracy: 0.5589 - val_loss: 1.2206 - val_accuracy: 0.5161\n",
      "Epoch 46/50\n",
      "6844/6844 [==============================] - 1s 143us/sample - loss: 1.1240 - accuracy: 0.5618 - val_loss: 1.2197 - val_accuracy: 0.5155\n",
      "Epoch 47/50\n",
      "6844/6844 [==============================] - 1s 129us/sample - loss: 1.1204 - accuracy: 0.5603 - val_loss: 1.2271 - val_accuracy: 0.5096\n",
      "Epoch 48/50\n",
      "6844/6844 [==============================] - 1s 175us/sample - loss: 1.1203 - accuracy: 0.5589 - val_loss: 1.2216 - val_accuracy: 0.5202\n",
      "Epoch 49/50\n",
      "6844/6844 [==============================] - 1s 152us/sample - loss: 1.1195 - accuracy: 0.5584 - val_loss: 1.2203 - val_accuracy: 0.5184\n",
      "Epoch 50/50\n",
      "6844/6844 [==============================] - 2s 221us/sample - loss: 1.1185 - accuracy: 0.5577 - val_loss: 1.2238 - val_accuracy: 0.5172\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(32, input_dim=46, activation='relu'))\n",
    "model.add(Dense(16, activation='relu'))\n",
    "model.add(Dense(7, activation='softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer=Adam(lr=0.001), metrics=['accuracy'])\n",
    "\n",
    "history=model.fit(train_x, train_y_one, epochs=50, batch_size=32, verbose=1,validation_data=(test_x,test_y_one))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:20:00.078292Z",
     "start_time": "2020-03-01T09:19:59.603501Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x576 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2) = plt.subplots(nrows = 1, ncols = 2, figsize = (20,8))\n",
    "\n",
    "ax_1=sns.lineplot(history.epoch,history.history['accuracy'],ax=ax1,label='Train')\n",
    "sns.lineplot(history.epoch,history.history['val_accuracy'],ax=ax1,label='Test')\n",
    "ax_1.set(xlabel='Epoch', ylabel='Accuracy')\n",
    "\n",
    "ax_2=sns.lineplot(history.epoch,history.history['loss'],ax=ax2,label='Train')\n",
    "sns.lineplot(history.epoch,history.history['val_loss'],ax=ax2,label='Test')\n",
    "ax_2.set(xlabel='Epoch', ylabel='Loss')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Config 3: changing optimizer, from ADAM to SGD, and we increase the learning rate and the batch size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:20:03.206902Z",
     "start_time": "2020-03-01T09:20:00.366560Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Falling back from v2 loop because of error: Failed to find data adapter that can handle input: <class 'pandas.core.frame.DataFrame'>, <class 'NoneType'>\n",
      "Train on 6844 samples, validate on 1711 samples\n",
      "Epoch 1/50\n",
      "6844/6844 [==============================] - 1s 119us/sample - loss: 1.8065 - accuracy: 0.2490 - val_loss: 1.6746 - val_accuracy: 0.3735\n",
      "Epoch 2/50\n",
      "6844/6844 [==============================] - 0s 9us/sample - loss: 1.6360 - accuracy: 0.3673 - val_loss: 1.5876 - val_accuracy: 0.3793\n",
      "Epoch 3/50\n",
      "6844/6844 [==============================] - 0s 8us/sample - loss: 1.5787 - accuracy: 0.3757 - val_loss: 1.5490 - val_accuracy: 0.3869\n",
      "Epoch 4/50\n",
      "6844/6844 [==============================] - 0s 6us/sample - loss: 1.5437 - accuracy: 0.3830 - val_loss: 1.5168 - val_accuracy: 0.3910\n",
      "Epoch 5/50\n",
      "6844/6844 [==============================] - 0s 8us/sample - loss: 1.5130 - accuracy: 0.3913 - val_loss: 1.4867 - val_accuracy: 0.3898\n",
      "Epoch 6/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.4833 - accuracy: 0.3960 - val_loss: 1.4576 - val_accuracy: 0.3945\n",
      "Epoch 7/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.4542 - accuracy: 0.4030 - val_loss: 1.4298 - val_accuracy: 0.4079\n",
      "Epoch 8/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.4260 - accuracy: 0.4125 - val_loss: 1.4036 - val_accuracy: 0.4337\n",
      "Epoch 9/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.4007 - accuracy: 0.4297 - val_loss: 1.3788 - val_accuracy: 0.4512\n",
      "Epoch 10/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.3751 - accuracy: 0.4459 - val_loss: 1.3569 - val_accuracy: 0.4600\n",
      "Epoch 11/50\n",
      "6844/6844 [==============================] - 0s 6us/sample - loss: 1.3527 - accuracy: 0.4601 - val_loss: 1.3377 - val_accuracy: 0.4728\n",
      "Epoch 12/50\n",
      "6844/6844 [==============================] - 0s 6us/sample - loss: 1.3324 - accuracy: 0.4722 - val_loss: 1.3238 - val_accuracy: 0.4863\n",
      "Epoch 13/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.3162 - accuracy: 0.4819 - val_loss: 1.3086 - val_accuracy: 0.4892\n",
      "Epoch 14/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.3013 - accuracy: 0.4909 - val_loss: 1.2964 - val_accuracy: 0.4991\n",
      "Epoch 15/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2888 - accuracy: 0.4985 - val_loss: 1.2841 - val_accuracy: 0.5038\n",
      "Epoch 16/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2784 - accuracy: 0.5082 - val_loss: 1.2759 - val_accuracy: 0.5143\n",
      "Epoch 17/50\n",
      "6844/6844 [==============================] - 0s 9us/sample - loss: 1.2691 - accuracy: 0.5187 - val_loss: 1.2694 - val_accuracy: 0.5102\n",
      "Epoch 18/50\n",
      "6844/6844 [==============================] - 0s 8us/sample - loss: 1.2635 - accuracy: 0.5186 - val_loss: 1.2648 - val_accuracy: 0.5137\n",
      "Epoch 19/50\n",
      "6844/6844 [==============================] - 0s 9us/sample - loss: 1.2567 - accuracy: 0.5231 - val_loss: 1.2590 - val_accuracy: 0.5237\n",
      "Epoch 20/50\n",
      "6844/6844 [==============================] - 0s 8us/sample - loss: 1.2512 - accuracy: 0.5314 - val_loss: 1.2578 - val_accuracy: 0.5178\n",
      "Epoch 21/50\n",
      "6844/6844 [==============================] - 0s 9us/sample - loss: 1.2475 - accuracy: 0.5314 - val_loss: 1.2512 - val_accuracy: 0.5225\n",
      "Epoch 22/50\n",
      "6844/6844 [==============================] - 0s 8us/sample - loss: 1.2440 - accuracy: 0.5335 - val_loss: 1.2476 - val_accuracy: 0.5184\n",
      "Epoch 23/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2403 - accuracy: 0.5359 - val_loss: 1.2444 - val_accuracy: 0.5213\n",
      "Epoch 24/50\n",
      "6844/6844 [==============================] - 0s 6us/sample - loss: 1.2373 - accuracy: 0.5365 - val_loss: 1.2418 - val_accuracy: 0.5307\n",
      "Epoch 25/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2345 - accuracy: 0.5390 - val_loss: 1.2400 - val_accuracy: 0.5231\n",
      "Epoch 26/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2322 - accuracy: 0.5355 - val_loss: 1.2373 - val_accuracy: 0.5243\n",
      "Epoch 27/50\n",
      "6844/6844 [==============================] - 0s 8us/sample - loss: 1.2302 - accuracy: 0.5370 - val_loss: 1.2351 - val_accuracy: 0.5260\n",
      "Epoch 28/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2285 - accuracy: 0.5381 - val_loss: 1.2320 - val_accuracy: 0.5225\n",
      "Epoch 29/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2279 - accuracy: 0.5381 - val_loss: 1.2302 - val_accuracy: 0.5266\n",
      "Epoch 30/50\n",
      "6844/6844 [==============================] - 0s 6us/sample - loss: 1.2250 - accuracy: 0.5414 - val_loss: 1.2304 - val_accuracy: 0.5301\n",
      "Epoch 31/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2252 - accuracy: 0.5414 - val_loss: 1.2276 - val_accuracy: 0.5330\n",
      "Epoch 32/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2246 - accuracy: 0.5395 - val_loss: 1.2269 - val_accuracy: 0.5301\n",
      "Epoch 33/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2209 - accuracy: 0.5414 - val_loss: 1.2268 - val_accuracy: 0.5342\n",
      "Epoch 34/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2193 - accuracy: 0.5418 - val_loss: 1.2267 - val_accuracy: 0.5307\n",
      "Epoch 35/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2182 - accuracy: 0.5418 - val_loss: 1.2226 - val_accuracy: 0.5377\n",
      "Epoch 36/50\n",
      "6844/6844 [==============================] - 0s 6us/sample - loss: 1.2180 - accuracy: 0.5408 - val_loss: 1.2238 - val_accuracy: 0.5295\n",
      "Epoch 37/50\n",
      "6844/6844 [==============================] - 0s 6us/sample - loss: 1.2155 - accuracy: 0.5450 - val_loss: 1.2217 - val_accuracy: 0.5301\n",
      "Epoch 38/50\n",
      "6844/6844 [==============================] - 0s 10us/sample - loss: 1.2154 - accuracy: 0.5432 - val_loss: 1.2252 - val_accuracy: 0.5307\n",
      "Epoch 39/50\n",
      "6844/6844 [==============================] - 0s 9us/sample - loss: 1.2134 - accuracy: 0.5427 - val_loss: 1.2211 - val_accuracy: 0.5336\n",
      "Epoch 40/50\n",
      "6844/6844 [==============================] - 0s 8us/sample - loss: 1.2137 - accuracy: 0.5431 - val_loss: 1.2199 - val_accuracy: 0.5336\n",
      "Epoch 41/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2126 - accuracy: 0.5470 - val_loss: 1.2196 - val_accuracy: 0.5330\n",
      "Epoch 42/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2107 - accuracy: 0.5468 - val_loss: 1.2185 - val_accuracy: 0.5359\n",
      "Epoch 43/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2106 - accuracy: 0.5432 - val_loss: 1.2173 - val_accuracy: 0.5330\n",
      "Epoch 44/50\n",
      "6844/6844 [==============================] - 0s 6us/sample - loss: 1.2100 - accuracy: 0.5415 - val_loss: 1.2176 - val_accuracy: 0.5342\n",
      "Epoch 45/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2091 - accuracy: 0.5469 - val_loss: 1.2190 - val_accuracy: 0.5354\n",
      "Epoch 46/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2100 - accuracy: 0.5437 - val_loss: 1.2174 - val_accuracy: 0.5371\n",
      "Epoch 47/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2090 - accuracy: 0.5430 - val_loss: 1.2155 - val_accuracy: 0.5348\n",
      "Epoch 48/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2071 - accuracy: 0.5438 - val_loss: 1.2161 - val_accuracy: 0.5383\n",
      "Epoch 49/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2067 - accuracy: 0.5430 - val_loss: 1.2151 - val_accuracy: 0.5400\n",
      "Epoch 50/50\n",
      "6844/6844 [==============================] - 0s 7us/sample - loss: 1.2062 - accuracy: 0.5463 - val_loss: 1.2165 - val_accuracy: 0.5377\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(32, input_dim=46, activation='relu'))\n",
    "model.add(Dense(16, activation='relu'))\n",
    "model.add(Dense(7, activation='softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer=SGD(lr=0.1,momentum=0.5,decay=0.1/50), metrics=['accuracy'])\n",
    "\n",
    "history=model.fit(train_x, train_y_one, epochs=50, batch_size=1000, verbose=1,validation_data=(test_x,test_y_one))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:20:03.941451Z",
     "start_time": "2020-03-01T09:20:03.477734Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x576 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2) = plt.subplots(nrows = 1, ncols = 2, figsize = (20,8))\n",
    "\n",
    "ax_1=sns.lineplot(history.epoch,history.history['accuracy'],ax=ax1,label='Train')\n",
    "sns.lineplot(history.epoch,history.history['val_accuracy'],ax=ax1,label='Test')\n",
    "ax_1.set(xlabel='Epoch', ylabel='Accuracy')\n",
    "\n",
    "ax_2=sns.lineplot(history.epoch,history.history['loss'],ax=ax2,label='Train')\n",
    "sns.lineplot(history.epoch,history.history['val_loss'],ax=ax2,label='Test')\n",
    "ax_2.set(xlabel='Epoch', ylabel='Loss')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The third configuration seems to be the more stable one since it is the one with less difference between the train and the test."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.4.4 Conclusions and predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Random forest and ANN seems to have more or less the same accuracy level (around 53%), with the ANN, in the last configuration, where it seems to be a little more precise. The level of accuracy is very low, but we need to consider that the algorithms were trained on a sample of around 6800 elements and tested on around 1800 elements. So, a big chunk of information has been cut in order to provide a meter of evaluation. Surely, by using all the possible information we will be able to gain a better amount of accuracy, despite the fact that we cannot test that.<br>\n",
    "We move on by generating the predictions on the last configuration of the ANN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:20:06.714052Z",
     "start_time": "2020-03-01T09:20:04.212282Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Falling back from v2 loop because of error: Failed to find data adapter that can handle input: <class 'pandas.core.frame.DataFrame'>, <class 'NoneType'>\n",
      "Train on 8555 samples\n",
      "Epoch 1/50\n",
      "8555/8555 [==============================] - 0s 9us/sample - loss: 1.1905 - accuracy: 0.5439\n",
      "Epoch 2/50\n",
      "8555/8555 [==============================] - 0s 7us/sample - loss: 1.1901 - accuracy: 0.5463\n",
      "Epoch 3/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1906 - accuracy: 0.5446\n",
      "Epoch 4/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1910 - accuracy: 0.5452\n",
      "Epoch 5/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1900 - accuracy: 0.5421\n",
      "Epoch 6/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1901 - accuracy: 0.5440\n",
      "Epoch 7/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1904 - accuracy: 0.5460\n",
      "Epoch 8/50\n",
      "8555/8555 [==============================] - 0s 5us/sample - loss: 1.1895 - accuracy: 0.5460\n",
      "Epoch 9/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1900 - accuracy: 0.5461\n",
      "Epoch 10/50\n",
      "8555/8555 [==============================] - 0s 5us/sample - loss: 1.1900 - accuracy: 0.5458\n",
      "Epoch 11/50\n",
      "8555/8555 [==============================] - 0s 7us/sample - loss: 1.1897 - accuracy: 0.5447\n",
      "Epoch 12/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1893 - accuracy: 0.5482\n",
      "Epoch 13/50\n",
      "8555/8555 [==============================] - 0s 7us/sample - loss: 1.1895 - accuracy: 0.5425\n",
      "Epoch 14/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1885 - accuracy: 0.5455\n",
      "Epoch 15/50\n",
      "8555/8555 [==============================] - 0s 5us/sample - loss: 1.1889 - accuracy: 0.5456\n",
      "Epoch 16/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1893 - accuracy: 0.5460\n",
      "Epoch 17/50\n",
      "8555/8555 [==============================] - 0s 5us/sample - loss: 1.1886 - accuracy: 0.5462\n",
      "Epoch 18/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1881 - accuracy: 0.5449\n",
      "Epoch 19/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1880 - accuracy: 0.5446\n",
      "Epoch 20/50\n",
      "8555/8555 [==============================] - 0s 7us/sample - loss: 1.1881 - accuracy: 0.5448\n",
      "Epoch 21/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1880 - accuracy: 0.5451\n",
      "Epoch 22/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1880 - accuracy: 0.5448\n",
      "Epoch 23/50\n",
      "8555/8555 [==============================] - 0s 7us/sample - loss: 1.1878 - accuracy: 0.5466\n",
      "Epoch 24/50\n",
      "8555/8555 [==============================] - 0s 8us/sample - loss: 1.1869 - accuracy: 0.5462\n",
      "Epoch 25/50\n",
      "8555/8555 [==============================] - 0s 8us/sample - loss: 1.1879 - accuracy: 0.5442\n",
      "Epoch 26/50\n",
      "8555/8555 [==============================] - 0s 7us/sample - loss: 1.1875 - accuracy: 0.5467\n",
      "Epoch 27/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1873 - accuracy: 0.5468\n",
      "Epoch 28/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1878 - accuracy: 0.5424\n",
      "Epoch 29/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1874 - accuracy: 0.5452\n",
      "Epoch 30/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1866 - accuracy: 0.5474\n",
      "Epoch 31/50\n",
      "8555/8555 [==============================] - 0s 8us/sample - loss: 1.1865 - accuracy: 0.5469\n",
      "Epoch 32/50\n",
      "8555/8555 [==============================] - 0s 9us/sample - loss: 1.1873 - accuracy: 0.5446\n",
      "Epoch 33/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1868 - accuracy: 0.5468\n",
      "Epoch 34/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1863 - accuracy: 0.5458\n",
      "Epoch 35/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1867 - accuracy: 0.5461\n",
      "Epoch 36/50\n",
      "8555/8555 [==============================] - 0s 8us/sample - loss: 1.1865 - accuracy: 0.5461\n",
      "Epoch 37/50\n",
      "8555/8555 [==============================] - 0s 8us/sample - loss: 1.1862 - accuracy: 0.5480\n",
      "Epoch 38/50\n",
      "8555/8555 [==============================] - 0s 8us/sample - loss: 1.1859 - accuracy: 0.5488\n",
      "Epoch 39/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1870 - accuracy: 0.5475\n",
      "Epoch 40/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1856 - accuracy: 0.5468\n",
      "Epoch 41/50\n",
      "8555/8555 [==============================] - 0s 5us/sample - loss: 1.1855 - accuracy: 0.5466\n",
      "Epoch 42/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1858 - accuracy: 0.5475\n",
      "Epoch 43/50\n",
      "8555/8555 [==============================] - 0s 5us/sample - loss: 1.1859 - accuracy: 0.5480\n",
      "Epoch 44/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1853 - accuracy: 0.5469\n",
      "Epoch 45/50\n",
      "8555/8555 [==============================] - 0s 5us/sample - loss: 1.1854 - accuracy: 0.5468\n",
      "Epoch 46/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1851 - accuracy: 0.5459\n",
      "Epoch 47/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1851 - accuracy: 0.5475\n",
      "Epoch 48/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1855 - accuracy: 0.5473\n",
      "Epoch 49/50\n",
      "8555/8555 [==============================] - 0s 6us/sample - loss: 1.1850 - accuracy: 0.5475\n",
      "Epoch 50/50\n",
      "8555/8555 [==============================] - 0s 5us/sample - loss: 1.1846 - accuracy: 0.5472\n",
      "WARNING:tensorflow:Falling back from v2 loop because of error: Failed to find data adapter that can handle input: <class 'pandas.core.frame.DataFrame'>, <class 'NoneType'>\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ANDREA\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:7: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  import sys\n"
     ]
    }
   ],
   "source": [
    "Y_dummy=pd.get_dummies(Y)\n",
    "\n",
    "history=model.fit(X,Y_dummy, epochs=50, batch_size=1000, verbose=1)\n",
    "\n",
    "X_imputation=val_ed[val_ed.columns.difference(['education'])]\n",
    "pred=model.predict_classes(X_imputation)\n",
    "val_ed['education']=list(pred)\n",
    "\n",
    "d=val_ed.append([sub_ed])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.5 Housing imputation through Random Forest"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have decided to perform the imputation on the variable \"Housing\" with a Random Forest, since the variable has less modalities and the structure of the Random Forest has less parameters to estimated. To do that, we performed again a k=5 fold cross validation on the train side. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:20:07.177767Z",
     "start_time": "2020-03-01T09:20:07.059839Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ANDREA\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:5: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  \"\"\"\n"
     ]
    }
   ],
   "source": [
    "data_house=d.merge(data_origin['housing'],on='ID')\n",
    "val_ho=data_house[data_house['housing']=='unknown']\n",
    "sub_ho=data_house[data_house['housing']!='unknown']\n",
    "\n",
    "sub_ho['housing']=pd.factorize(sub_ho['housing'])[0]\n",
    "\n",
    "X=sub_ho[sub_ho.columns.difference(['housing'])]\n",
    "Y=sub_ho['housing']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.5.1 Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:25:19.703639Z",
     "start_time": "2020-03-01T09:20:07.581518Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "50   0.5650438946528332   {'min_samples_leaf': 0.1, 'max_features': 0.5, 'max_depth': 2, 'criterion': 'gini'}\n",
      "100   0.5641317979705849   {'min_samples_leaf': 0.1, 'max_features': 0.5, 'max_depth': 2, 'criterion': 'entropy'}\n",
      "150   0.5639037738000228   {'min_samples_leaf': 0.1, 'max_features': 0.5, 'max_depth': 3, 'criterion': 'entropy'}\n",
      "200   0.5641317979705849   {'min_samples_leaf': 0.1, 'max_features': 0.5, 'max_depth': 2, 'criterion': 'gini'}\n"
     ]
    }
   ],
   "source": [
    "for i in [50,100,150,200]:\n",
    "    rf = RandomForestClassifier(random_state=33, n_estimators=i)\n",
    "    params = {'criterion': ['gini', 'entropy'], 'max_depth': np.arange(2, 6), 'min_samples_leaf': np.logspace(-1, -3, 3),'max_features': ['sqrt', 0.2, 0.5]}\n",
    "    cv_tuning = RandomizedSearchCV(rf, params, random_state=122223, scoring='accuracy', return_train_score=True, cv=5, n_iter=20)\n",
    "    cv_tuning.fit(X,Y)\n",
    "    print(i,' ',cv_tuning.best_score_,' ',cv_tuning.best_params_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The best model seems to be the first configuration with 50 trees. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:25:20.443026Z",
     "start_time": "2020-03-01T09:25:20.122224Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "                       max_depth=2, max_features=0.5, max_leaf_nodes=None,\n",
       "                       min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                       min_samples_leaf=0.1, min_samples_split=2,\n",
       "                       min_weight_fraction_leaf=0.0, n_estimators=50,\n",
       "                       n_jobs=None, oob_score=False, random_state=33, verbose=0,\n",
       "                       warm_start=False)"
      ]
     },
     "execution_count": 99,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf = RandomForestClassifier(random_state=33, n_estimators=50,\n",
    "                            min_samples_leaf=0.1,\n",
    "                            max_features=0.5,\n",
    "                            max_depth=2,\n",
    "                            criterion='gini')\n",
    "rf.fit(X,Y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.5.2 Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:25:21.027145Z",
     "start_time": "2020-03-01T09:25:20.906220Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ANDREA\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:5: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  \"\"\"\n"
     ]
    }
   ],
   "source": [
    "X_imputation=val_ho[val_ho.columns.difference(['housing'])]\n",
    "\n",
    "predictions = rf.predict(X_imputation)\n",
    "\n",
    "val_ho['housing']=list(predictions)\n",
    "\n",
    "d=val_ho.append([sub_ho])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.6 Loan imputation through Random Forest"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Loan has the same structure as Housing. So, we decided to go with the same approach as for Housing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:25:21.553414Z",
     "start_time": "2020-03-01T09:25:21.425492Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ANDREA\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:6: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  \n"
     ]
    }
   ],
   "source": [
    "data_loan=d.merge(data_origin['loan'],on='ID')\n",
    "\n",
    "val_lo=data_loan[data_loan['loan']=='unknown']\n",
    "sub_lo=data_loan[data_loan['loan']!='unknown']\n",
    "\n",
    "sub_lo['loan']=pd.factorize(sub_lo['loan'])[0]\n",
    "\n",
    "X=sub_lo[sub_lo.columns.difference(['loan'])]\n",
    "Y=sub_lo['loan']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.6.1 Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:28:53.737183Z",
     "start_time": "2020-03-01T09:25:22.022126Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "50   0.8483639265762171   {'min_samples_leaf': 0.001, 'max_features': 'sqrt', 'max_depth': 2, 'criterion': 'gini'}\n",
      "100   0.8483639265762171   {'min_samples_leaf': 0.001, 'max_features': 'sqrt', 'max_depth': 2, 'criterion': 'gini'}\n",
      "150   0.8483639265762171   {'min_samples_leaf': 0.001, 'max_features': 'sqrt', 'max_depth': 2, 'criterion': 'gini'}\n",
      "200   0.8483639265762171   {'min_samples_leaf': 0.001, 'max_features': 'sqrt', 'max_depth': 2, 'criterion': 'gini'}\n"
     ]
    }
   ],
   "source": [
    "for i in [50,100,150,200]:\n",
    "    rf = RandomForestClassifier(random_state=33, n_estimators=i)\n",
    "    params = {'criterion': ['gini', 'entropy'], 'max_depth': np.arange(2, 6), 'min_samples_leaf': np.logspace(-1, -3, 3),'max_features': ['sqrt', 0.2, 0.5]}\n",
    "    cv_tuning = RandomizedSearchCV(rf, params, random_state=122223, scoring='accuracy', return_train_score=True, cv=5, n_iter=20)\n",
    "    cv_tuning.fit(X,Y)\n",
    "    print(i,' ',cv_tuning.best_score_,' ',cv_tuning.best_params_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Every random forest has the same amount of accuracy, the increase in the number of trees that are generated each time does not influence the accuracy. We will go with the easiest configuration with only 50 trees."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:28:54.475730Z",
     "start_time": "2020-03-01T09:28:54.283846Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "                       max_depth=2, max_features='sqrt', max_leaf_nodes=None,\n",
       "                       min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                       min_samples_leaf=0.001, min_samples_split=2,\n",
       "                       min_weight_fraction_leaf=0.0, n_estimators=50,\n",
       "                       n_jobs=None, oob_score=False, random_state=33, verbose=0,\n",
       "                       warm_start=False)"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf = RandomForestClassifier(random_state=33, n_estimators=50, min_samples_leaf=0.001,max_features='sqrt',max_depth=2,criterion='gini')\n",
    "rf.fit(X,Y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.6.2 Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:28:55.128411Z",
     "start_time": "2020-03-01T09:28:55.007400Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ANDREA\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:5: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  \"\"\"\n"
     ]
    }
   ],
   "source": [
    "X_imputation=val_lo[val_lo.columns.difference(['loan'])]\n",
    "\n",
    "predictions = rf.predict(X_imputation)\n",
    "\n",
    "val_lo['loan']=list(predictions)\n",
    "d=val_lo.append([sub_lo])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.6.3 Final considerations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We think that the strange behaviour  of the accuracy in the Random Forest is justified by the amount of \"No\" (0s in the reparametrization).This leads to the fact that the performances of the random forests are more or less the same of an imputation through a mode. This can be proved by looking at the amount of probability that the RF assigns to the first class prediction."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:28:55.697060Z",
     "start_time": "2020-03-01T09:28:55.672076Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.84480326, 0.15519674],\n",
       "       [0.84591765, 0.15408235],\n",
       "       [0.82244208, 0.17755792],\n",
       "       [0.85316991, 0.14683009],\n",
       "       [0.85311665, 0.14688335],\n",
       "       [0.83152608, 0.16847392],\n",
       "       [0.8382006 , 0.1617994 ],\n",
       "       [0.84127283, 0.15872717],\n",
       "       [0.80839525, 0.19160475],\n",
       "       [0.83784071, 0.16215929],\n",
       "       [0.85418805, 0.14581195],\n",
       "       [0.84823159, 0.15176841],\n",
       "       [0.84456709, 0.15543291],\n",
       "       [0.82423648, 0.17576352],\n",
       "       [0.84847281, 0.15152719],\n",
       "       [0.84731122, 0.15268878],\n",
       "       [0.8533499 , 0.1466501 ],\n",
       "       [0.85204652, 0.14795348],\n",
       "       [0.85209128, 0.14790872],\n",
       "       [0.84991226, 0.15008774],\n",
       "       [0.8408443 , 0.1591557 ],\n",
       "       [0.85171831, 0.14828169],\n",
       "       [0.8478479 , 0.1521521 ],\n",
       "       [0.85414677, 0.14585323],\n",
       "       [0.85117796, 0.14882204],\n",
       "       [0.84792041, 0.15207959],\n",
       "       [0.85366987, 0.14633013],\n",
       "       [0.85462995, 0.14537005],\n",
       "       [0.84962269, 0.15037731],\n",
       "       [0.84632448, 0.15367552],\n",
       "       [0.84901711, 0.15098289],\n",
       "       [0.85120774, 0.14879226],\n",
       "       [0.85422659, 0.14577341],\n",
       "       [0.85000346, 0.14999654],\n",
       "       [0.85276582, 0.14723418],\n",
       "       [0.85462738, 0.14537262],\n",
       "       [0.85040074, 0.14959926],\n",
       "       [0.85336087, 0.14663913],\n",
       "       [0.85390318, 0.14609682],\n",
       "       [0.84784147, 0.15215853],\n",
       "       [0.85153054, 0.14846946],\n",
       "       [0.8539938 , 0.1460062 ],\n",
       "       [0.84978265, 0.15021735],\n",
       "       [0.84327151, 0.15672849],\n",
       "       [0.84714243, 0.15285757],\n",
       "       [0.86297278, 0.13702722],\n",
       "       [0.84532059, 0.15467941],\n",
       "       [0.84934345, 0.15065655],\n",
       "       [0.85414677, 0.14585323],\n",
       "       [0.85173619, 0.14826381],\n",
       "       [0.8474939 , 0.1525061 ],\n",
       "       [0.85157221, 0.14842779],\n",
       "       [0.84899578, 0.15100422],\n",
       "       [0.8509337 , 0.1490663 ],\n",
       "       [0.85349826, 0.14650174],\n",
       "       [0.84693667, 0.15306333],\n",
       "       [0.85405984, 0.14594016],\n",
       "       [0.84677499, 0.15322501],\n",
       "       [0.84816319, 0.15183681],\n",
       "       [0.84549361, 0.15450639],\n",
       "       [0.83918263, 0.16081737],\n",
       "       [0.8484663 , 0.1515337 ],\n",
       "       [0.85240275, 0.14759725],\n",
       "       [0.84454287, 0.15545713],\n",
       "       [0.85462738, 0.14537262],\n",
       "       [0.84227478, 0.15772522],\n",
       "       [0.85066741, 0.14933259],\n",
       "       [0.84653506, 0.15346494],\n",
       "       [0.85262409, 0.14737591],\n",
       "       [0.85249898, 0.14750102],\n",
       "       [0.84713888, 0.15286112],\n",
       "       [0.84466544, 0.15533456],\n",
       "       [0.85183909, 0.14816091],\n",
       "       [0.83441206, 0.16558794],\n",
       "       [0.84757396, 0.15242604],\n",
       "       [0.84658297, 0.15341703],\n",
       "       [0.8260798 , 0.1739202 ],\n",
       "       [0.84741952, 0.15258048],\n",
       "       [0.85319321, 0.14680679],\n",
       "       [0.85187752, 0.14812248],\n",
       "       [0.85118244, 0.14881756],\n",
       "       [0.84509755, 0.15490245],\n",
       "       [0.84727787, 0.15272213],\n",
       "       [0.84789382, 0.15210618],\n",
       "       [0.85468822, 0.14531178],\n",
       "       [0.85127579, 0.14872421],\n",
       "       [0.84909378, 0.15090622],\n",
       "       [0.85130045, 0.14869955],\n",
       "       [0.85227871, 0.14772129],\n",
       "       [0.84527656, 0.15472344],\n",
       "       [0.84231707, 0.15768293],\n",
       "       [0.86414658, 0.13585342],\n",
       "       [0.85134937, 0.14865063],\n",
       "       [0.8528569 , 0.1471431 ],\n",
       "       [0.85592489, 0.14407511],\n",
       "       [0.85110459, 0.14889541],\n",
       "       [0.85005247, 0.14994753],\n",
       "       [0.85462593, 0.14537407],\n",
       "       [0.84579361, 0.15420639],\n",
       "       [0.85246142, 0.14753858],\n",
       "       [0.84747545, 0.15252455],\n",
       "       [0.84535645, 0.15464355],\n",
       "       [0.84761309, 0.15238691],\n",
       "       [0.85187129, 0.14812871],\n",
       "       [0.85298437, 0.14701563],\n",
       "       [0.85216024, 0.14783976],\n",
       "       [0.8538035 , 0.1461965 ],\n",
       "       [0.85277861, 0.14722139],\n",
       "       [0.85414677, 0.14585323],\n",
       "       [0.84637812, 0.15362188],\n",
       "       [0.84572439, 0.15427561],\n",
       "       [0.84716049, 0.15283951],\n",
       "       [0.84741952, 0.15258048],\n",
       "       [0.8495249 , 0.1504751 ],\n",
       "       [0.84915703, 0.15084297],\n",
       "       [0.84825532, 0.15174468],\n",
       "       [0.84914788, 0.15085212],\n",
       "       [0.84869261, 0.15130739],\n",
       "       [0.85036271, 0.14963729],\n",
       "       [0.8536553 , 0.1463447 ],\n",
       "       [0.84163196, 0.15836804],\n",
       "       [0.85037037, 0.14962963],\n",
       "       [0.84754678, 0.15245322],\n",
       "       [0.84584284, 0.15415716],\n",
       "       [0.84757789, 0.15242211],\n",
       "       [0.84794463, 0.15205537],\n",
       "       [0.85354185, 0.14645815],\n",
       "       [0.84588969, 0.15411031],\n",
       "       [0.85328748, 0.14671252],\n",
       "       [0.84950899, 0.15049101],\n",
       "       [0.84707613, 0.15292387],\n",
       "       [0.84910202, 0.15089798],\n",
       "       [0.84589188, 0.15410812],\n",
       "       [0.84642411, 0.15357589],\n",
       "       [0.85550035, 0.14449965],\n",
       "       [0.84520192, 0.15479808],\n",
       "       [0.85315656, 0.14684344],\n",
       "       [0.84642283, 0.15357717],\n",
       "       [0.84639871, 0.15360129],\n",
       "       [0.8491373 , 0.1508627 ],\n",
       "       [0.84694441, 0.15305559],\n",
       "       [0.84473032, 0.15526968],\n",
       "       [0.84190064, 0.15809936],\n",
       "       [0.84859157, 0.15140843],\n",
       "       [0.84602401, 0.15397599],\n",
       "       [0.84901711, 0.15098289],\n",
       "       [0.84933068, 0.15066932],\n",
       "       [0.84628903, 0.15371097],\n",
       "       [0.84878337, 0.15121663],\n",
       "       [0.84865623, 0.15134377],\n",
       "       [0.84660037, 0.15339963],\n",
       "       [0.84360612, 0.15639388],\n",
       "       [0.84553456, 0.15446544],\n",
       "       [0.83881819, 0.16118181],\n",
       "       [0.85333189, 0.14666811],\n",
       "       [0.86427493, 0.13572507],\n",
       "       [0.85245469, 0.14754531],\n",
       "       [0.84741467, 0.15258533],\n",
       "       [0.84690913, 0.15309087],\n",
       "       [0.84738319, 0.15261681],\n",
       "       [0.84880051, 0.15119949],\n",
       "       [0.84427902, 0.15572098],\n",
       "       [0.85263798, 0.14736202],\n",
       "       [0.8518784 , 0.1481216 ],\n",
       "       [0.85951367, 0.14048633],\n",
       "       [0.84563569, 0.15436431],\n",
       "       [0.84017735, 0.15982265],\n",
       "       [0.84494836, 0.15505164],\n",
       "       [0.84954754, 0.15045246],\n",
       "       [0.84859257, 0.15140743],\n",
       "       [0.8650781 , 0.1349219 ],\n",
       "       [0.8484068 , 0.1515932 ],\n",
       "       [0.84462262, 0.15537738],\n",
       "       [0.84833486, 0.15166514],\n",
       "       [0.851838  , 0.148162  ],\n",
       "       [0.83623293, 0.16376707],\n",
       "       [0.83956764, 0.16043236],\n",
       "       [0.84770231, 0.15229769],\n",
       "       [0.84331364, 0.15668636],\n",
       "       [0.85254142, 0.14745858],\n",
       "       [0.83855602, 0.16144398],\n",
       "       [0.85043352, 0.14956648],\n",
       "       [0.84742473, 0.15257527],\n",
       "       [0.85262409, 0.14737591],\n",
       "       [0.84518292, 0.15481708],\n",
       "       [0.8559267 , 0.1440733 ],\n",
       "       [0.84530406, 0.15469594],\n",
       "       [0.85013661, 0.14986339],\n",
       "       [0.84969211, 0.15030789],\n",
       "       [0.8491816 , 0.1508184 ],\n",
       "       [0.85199326, 0.14800674],\n",
       "       [0.81570213, 0.18429787],\n",
       "       [0.85405464, 0.14594536],\n",
       "       [0.85428888, 0.14571112],\n",
       "       [0.85456509, 0.14543491],\n",
       "       [0.85289034, 0.14710966],\n",
       "       [0.8489206 , 0.1510794 ],\n",
       "       [0.85049833, 0.14950167],\n",
       "       [0.85276582, 0.14723418],\n",
       "       [0.85476971, 0.14523029],\n",
       "       [0.85134317, 0.14865683],\n",
       "       [0.84894929, 0.15105071],\n",
       "       [0.85478035, 0.14521965],\n",
       "       [0.85503249, 0.14496751],\n",
       "       [0.85531251, 0.14468749],\n",
       "       [0.85161224, 0.14838776],\n",
       "       [0.84970759, 0.15029241],\n",
       "       [0.84576068, 0.15423932],\n",
       "       [0.84677738, 0.15322262]])"
      ]
     },
     "execution_count": 104,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predi=rf.predict_proba(X_imputation)\n",
    "predi"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It is possible to see that the first (0) class has always a higher probability than the second one (1). Moreover, if we look at the value counts of the dataset before and after the imputation, we can see that all the `unknown` are placed inside the 0s class (like a mode). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:28:56.180765Z",
     "start_time": "2020-03-01T09:28:56.169770Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original dataset\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "no         7441\n",
       "yes        1330\n",
       "unknown     209\n",
       "Name: loan, dtype: int64"
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Original dataset')\n",
    "data.loan.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:28:56.692299Z",
     "start_time": "2020-03-01T09:28:56.680305Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0    7650\n",
       "1    1330\n",
       "Name: loan, dtype: int64"
      ]
     },
     "execution_count": 106,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('')\n",
    "d.loan.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.7 Saving the new dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-03-01T09:28:57.289411Z",
     "start_time": "2020-03-01T09:28:57.243441Z"
    }
   },
   "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>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>campaign</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "      <th>y</th>\n",
       "      <th>pdays_new</th>\n",
       "      <th>education</th>\n",
       "      <th>loan</th>\n",
       "      <th>housing</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ID</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>34579</th>\n",
       "      <td>35</td>\n",
       "      <td>admin.</td>\n",
       "      <td>single</td>\n",
       "      <td>cellular</td>\n",
       "      <td>may</td>\n",
       "      <td>thu</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>failure</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.266</td>\n",
       "      <td>5099.1</td>\n",
       "      <td>no</td>\n",
       "      <td>not_rec_cont</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>446</th>\n",
       "      <td>42</td>\n",
       "      <td>technician</td>\n",
       "      <td>married</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>tue</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "      <td>yes</td>\n",
       "      <td>not_rec_cont</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20173</th>\n",
       "      <td>36</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>cellular</td>\n",
       "      <td>aug</td>\n",
       "      <td>mon</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.444</td>\n",
       "      <td>-36.1</td>\n",
       "      <td>4.965</td>\n",
       "      <td>5228.1</td>\n",
       "      <td>yes</td>\n",
       "      <td>not_rec_cont</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18171</th>\n",
       "      <td>37</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>telephone</td>\n",
       "      <td>jul</td>\n",
       "      <td>wed</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.918</td>\n",
       "      <td>-42.7</td>\n",
       "      <td>4.963</td>\n",
       "      <td>5228.1</td>\n",
       "      <td>yes</td>\n",
       "      <td>not_rec_cont</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30128</th>\n",
       "      <td>31</td>\n",
       "      <td>management</td>\n",
       "      <td>single</td>\n",
       "      <td>cellular</td>\n",
       "      <td>apr</td>\n",
       "      <td>thu</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>93.075</td>\n",
       "      <td>-47.1</td>\n",
       "      <td>1.365</td>\n",
       "      <td>5099.1</td>\n",
       "      <td>no</td>\n",
       "      <td>not_rec_cont</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       age         job  marital    contact month day_of_week  campaign  \\\n",
       "ID                                                                       \n",
       "34579   35      admin.   single   cellular   may         thu         1   \n",
       "446     42  technician  married  telephone   may         tue         1   \n",
       "20173   36      admin.  married   cellular   aug         mon         2   \n",
       "18171   37      admin.  married  telephone   jul         wed         2   \n",
       "30128   31  management   single   cellular   apr         thu         1   \n",
       "\n",
       "       previous     poutcome  emp.var.rate  cons.price.idx  cons.conf.idx  \\\n",
       "ID                                                                          \n",
       "34579         1      failure          -1.8          92.893          -46.2   \n",
       "446           0  nonexistent           1.1          93.994          -36.4   \n",
       "20173         0  nonexistent           1.4          93.444          -36.1   \n",
       "18171         0  nonexistent           1.4          93.918          -42.7   \n",
       "30128         0  nonexistent          -1.8          93.075          -47.1   \n",
       "\n",
       "       euribor3m  nr.employed    y     pdays_new  education  loan  housing  \n",
       "ID                                                                          \n",
       "34579      1.266       5099.1   no  not_rec_cont          6     0        1  \n",
       "446        4.857       5191.0  yes  not_rec_cont          5     0        0  \n",
       "20173      4.965       5228.1  yes  not_rec_cont          6     0        0  \n",
       "18171      4.963       5228.1  yes  not_rec_cont          4     1        1  \n",
       "30128      1.365       5099.1   no  not_rec_cont          6     0        1  "
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=data.drop(['education','loan','housing'],axis=1)\n",
    "dataset=data.merge(d[['education','loan','housing']],on='ID')\n",
    "#dataset_end=dataset.to_csv(r'C:\\Users\\ANDREA\\Desktop\\Data mining\\Dataset_end.csv', index = True)\n",
    "dataset.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}