Oil Trend Prediction .ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Project : Prediction of the Crude Oil production trend\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## l. Import the libraries \n",
    "\n",
    "load all useful python modules :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/benjelloun/anaconda/lib/python3.6/site-packages/sklearn/cross_validation.py:44: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n",
      "  \"This module will be removed in 0.20.\", DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "# Basic import numpy, pandas and matplotlib :\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy.random as npr\n",
    "import matplotlib.cm as cm\n",
    "\n",
    "# dimentionality reduction imports :\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.manifold import TSNE\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import KFold\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# useful build-in python functions\n",
    "from sklearn import metrics\n",
    "from sklearn.pipeline import Pipeline\n",
    "\n",
    "\n",
    "#Import classification algorithms :\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn import svm\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.ensemble import ExtraTreesClassifier\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from xgboost import XGBClassifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ll. Get the dataset\n",
    "\n",
    "We load the dataset from the given files (link to the computer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#_______ Links to the data \n",
    "\n",
    "# link to Train inputs :\n",
    "link_Train_input  = '/Users/benjelloun/Desktop/Challenge/Train_Input.csv'\n",
    "# link to Train output :\n",
    "link_Train_output = '/Users/benjelloun/Desktop/Challenge/Training_Output.csv'\n",
    "# link to Test inputs :\n",
    "link_Test_input   = '/Users/benjelloun/Desktop/Challenge/Test.csv'\n",
    "\n",
    "#_______ Open the files \n",
    "#Train_Outputs :\n",
    "Train_Outputs = pd.read_csv(link_Train_output, sep = \";\")\n",
    "#Train_Inputs :\n",
    "Train_Input = pd.read_csv(link_Train_input, sep = \";\")\n",
    "#Test Input :\n",
    "Test_Input = pd.read_csv(link_Test_input, sep = \";\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## III. Preprocessing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Hands on data :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input Train data shape is : (10159, 123)\n",
      "Output Train data shape is : (10159, 2)\n",
      "the types present in the dataset : [dtype('int64') dtype('float64')]\n"
     ]
    }
   ],
   "source": [
    "#Get train input data shape\n",
    "print('Input Train data shape is :', Train_Input.shape)\n",
    "#Get train output data shape\n",
    "print('Output Train data shape is :', Train_Outputs.shape)\n",
    "#Get the types present in the data\n",
    "print('the types present in the dataset :', pd.unique( Train_Input.iloc[:,1:].dtypes ) )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ID                                ID00001\n",
      "month                                   5\n",
      "country                                52\n",
      "1_diffClosing stocks(kmt)               0\n",
      "1_diffExports(kmt)                    237\n",
      "1_diffImports(kmt)                      0\n",
      "1_diffRefinery intake(kmt)             33\n",
      "1_diffWTI                           -5.62\n",
      "1_diffSumClosing stocks(kmt)      10426.2\n",
      "1_diffSumExports(kmt)             12135.1\n",
      "1_diffSumImports(kmt)             20527.8\n",
      "1_diffSumProduction(kmt)          32234.6\n",
      "1_diffSumRefinery intake(kmt)     25998.8\n",
      "2_diffClosing stocks(kmt)               0\n",
      "2_diffExports(kmt)                   -159\n",
      "2_diffImports(kmt)                      0\n",
      "2_diffRefinery intake(kmt)            -17\n",
      "2_diffWTI                           -5.05\n",
      "2_diffSumClosing stocks(kmt)        -3510\n",
      "2_diffSumExports(kmt)                1124\n",
      "2_diffSumImports(kmt)            -6034.04\n",
      "2_diffSumProduction(kmt)          -2105.3\n",
      "2_diffSumRefinery intake(kmt)    -7860.37\n",
      "3_diffClosing stocks(kmt)               0\n",
      "3_diffExports(kmt)                     66\n",
      "3_diffImports(kmt)                      0\n",
      "3_diffRefinery intake(kmt)             12\n",
      "3_diffWTI                            3.47\n",
      "3_diffSumClosing stocks(kmt)     -1389.79\n",
      "3_diffSumExports(kmt)             8409.35\n",
      "                                   ...   \n",
      "10_diffClosing stocks(kmt)              0\n",
      "10_diffExports(kmt)                   317\n",
      "10_diffImports(kmt)                     0\n",
      "10_diffRefinery intake(kmt)            20\n",
      "10_diffWTI                           2.18\n",
      "10_diffSumClosing stocks(kmt)     1148.76\n",
      "10_diffSumExports(kmt)            9992.29\n",
      "10_diffSumImports(kmt)            11136.5\n",
      "10_diffSumProduction(kmt)           12455\n",
      "10_diffSumRefinery intake(kmt)    9752.46\n",
      "11_diffClosing stocks(kmt)              0\n",
      "11_diffExports(kmt)                     0\n",
      "11_diffImports(kmt)                     0\n",
      "11_diffRefinery intake(kmt)           -39\n",
      "11_diffWTI                           0.65\n",
      "11_diffSumClosing stocks(kmt)     11955.6\n",
      "11_diffSumExports(kmt)            2183.35\n",
      "11_diffSumImports(kmt)            2914.96\n",
      "11_diffSumProduction(kmt)        -258.389\n",
      "11_diffSumRefinery intake(kmt)    4412.76\n",
      "12_diffClosing stocks(kmt)              0\n",
      "12_diffExports(kmt)                     0\n",
      "12_diffImports(kmt)                     0\n",
      "12_diffRefinery intake(kmt)            14\n",
      "12_diffWTI                           2.92\n",
      "12_diffSumClosing stocks(kmt)     673.857\n",
      "12_diffSumExports(kmt)           -13007.2\n",
      "12_diffSumImports(kmt)           -5932.55\n",
      "12_diffSumProduction(kmt)        -17130.8\n",
      "12_diffSumRefinery intake(kmt)   -13151.5\n",
      "Name: 0, dtype: object\n"
     ]
    }
   ],
   "source": [
    "# Get a first idea of what looks the data like \n",
    "print( Train_Input.iloc[0] )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11632beb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## number of data per month :\n",
    "\n",
    "#get the statistic\n",
    "data_by_month = []\n",
    "for i in range(1,13):\n",
    "    data_by_month.append([i, np.sum(Train_Input.iloc[:, 1]==i)] )\n",
    "data_by_month = np.array(data_by_month)\n",
    "\n",
    "#bar plot :\n",
    "fig1 = plt.figure()\n",
    "plt.bar(data_by_month[:,0], data_by_month[:,1], align='center')\n",
    "plt.ylabel('Number of data')\n",
    "plt.xlabel('months')\n",
    "plt.title('Number of data per month')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of countries is  76\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQi4QZmZW6P8DFhTWXGfDTZoAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115a9b940>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## number of data per country\n",
    "#number of countries :\n",
    "number_of_countries = len(np.unique(Train_Input.iloc[:, 2]))\n",
    "print('number of countries is ', number_of_countries)\n",
    "\n",
    "#Get the statistics\n",
    "data_per_country = []\n",
    "for i in range(1, number_of_countries):\n",
    "    data_per_country.append([i, np.sum(Train_Input.iloc[:, 2]==i) ])\n",
    "data_per_country = np.array(data_per_country)\n",
    "\n",
    "# bar plot :\n",
    "fig2 = plt.figure()\n",
    "plt.bar(data_per_country[:,0], data_per_country[:,1], align='center')\n",
    "plt.ylabel('Number of data')\n",
    "plt.xlabel('country number')\n",
    "plt.title('Number of data per country')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['ID',\n",
       " 'month',\n",
       " 'country',\n",
       " '1_diffClosing stocks(kmt)',\n",
       " '1_diffExports(kmt)',\n",
       " '1_diffImports(kmt)',\n",
       " '1_diffRefinery intake(kmt)',\n",
       " '1_diffWTI',\n",
       " '1_diffSumClosing stocks(kmt)',\n",
       " '1_diffSumExports(kmt)',\n",
       " '1_diffSumImports(kmt)',\n",
       " '1_diffSumProduction(kmt)',\n",
       " '1_diffSumRefinery intake(kmt)',\n",
       " '2_diffClosing stocks(kmt)',\n",
       " '2_diffExports(kmt)',\n",
       " '2_diffImports(kmt)',\n",
       " '2_diffRefinery intake(kmt)',\n",
       " '2_diffWTI',\n",
       " '2_diffSumClosing stocks(kmt)',\n",
       " '2_diffSumExports(kmt)',\n",
       " '2_diffSumImports(kmt)',\n",
       " '2_diffSumProduction(kmt)',\n",
       " '2_diffSumRefinery intake(kmt)',\n",
       " '3_diffClosing stocks(kmt)',\n",
       " '3_diffExports(kmt)',\n",
       " '3_diffImports(kmt)',\n",
       " '3_diffRefinery intake(kmt)',\n",
       " '3_diffWTI',\n",
       " '3_diffSumClosing stocks(kmt)',\n",
       " '3_diffSumExports(kmt)',\n",
       " '3_diffSumImports(kmt)',\n",
       " '3_diffSumProduction(kmt)',\n",
       " '3_diffSumRefinery intake(kmt)',\n",
       " '4_diffClosing stocks(kmt)',\n",
       " '4_diffExports(kmt)',\n",
       " '4_diffImports(kmt)',\n",
       " '4_diffRefinery intake(kmt)',\n",
       " '4_diffWTI',\n",
       " '4_diffSumClosing stocks(kmt)',\n",
       " '4_diffSumExports(kmt)',\n",
       " '4_diffSumImports(kmt)',\n",
       " '4_diffSumProduction(kmt)',\n",
       " '4_diffSumRefinery intake(kmt)',\n",
       " '5_diffClosing stocks(kmt)',\n",
       " '5_diffExports(kmt)',\n",
       " '5_diffImports(kmt)',\n",
       " '5_diffRefinery intake(kmt)',\n",
       " '5_diffWTI',\n",
       " '5_diffSumClosing stocks(kmt)',\n",
       " '5_diffSumExports(kmt)',\n",
       " '5_diffSumImports(kmt)',\n",
       " '5_diffSumProduction(kmt)',\n",
       " '5_diffSumRefinery intake(kmt)',\n",
       " '6_diffClosing stocks(kmt)',\n",
       " '6_diffExports(kmt)',\n",
       " '6_diffImports(kmt)',\n",
       " '6_diffRefinery intake(kmt)',\n",
       " '6_diffWTI',\n",
       " '6_diffSumClosing stocks(kmt)',\n",
       " '6_diffSumExports(kmt)',\n",
       " '6_diffSumImports(kmt)',\n",
       " '6_diffSumProduction(kmt)',\n",
       " '6_diffSumRefinery intake(kmt)',\n",
       " '7_diffClosing stocks(kmt)',\n",
       " '7_diffExports(kmt)',\n",
       " '7_diffImports(kmt)',\n",
       " '7_diffRefinery intake(kmt)',\n",
       " '7_diffWTI',\n",
       " '7_diffSumClosing stocks(kmt)',\n",
       " '7_diffSumExports(kmt)',\n",
       " '7_diffSumImports(kmt)',\n",
       " '7_diffSumProduction(kmt)',\n",
       " '7_diffSumRefinery intake(kmt)',\n",
       " '8_diffClosing stocks(kmt)',\n",
       " '8_diffExports(kmt)',\n",
       " '8_diffImports(kmt)',\n",
       " '8_diffRefinery intake(kmt)',\n",
       " '8_diffWTI',\n",
       " '8_diffSumClosing stocks(kmt)',\n",
       " '8_diffSumExports(kmt)',\n",
       " '8_diffSumImports(kmt)',\n",
       " '8_diffSumProduction(kmt)',\n",
       " '8_diffSumRefinery intake(kmt)',\n",
       " '9_diffClosing stocks(kmt)',\n",
       " '9_diffExports(kmt)',\n",
       " '9_diffImports(kmt)',\n",
       " '9_diffRefinery intake(kmt)',\n",
       " '9_diffWTI',\n",
       " '9_diffSumClosing stocks(kmt)',\n",
       " '9_diffSumExports(kmt)',\n",
       " '9_diffSumImports(kmt)',\n",
       " '9_diffSumProduction(kmt)',\n",
       " '9_diffSumRefinery intake(kmt)',\n",
       " '10_diffClosing stocks(kmt)',\n",
       " '10_diffExports(kmt)',\n",
       " '10_diffImports(kmt)',\n",
       " '10_diffRefinery intake(kmt)',\n",
       " '10_diffWTI',\n",
       " '10_diffSumClosing stocks(kmt)',\n",
       " '10_diffSumExports(kmt)',\n",
       " '10_diffSumImports(kmt)',\n",
       " '10_diffSumProduction(kmt)',\n",
       " '10_diffSumRefinery intake(kmt)',\n",
       " '11_diffClosing stocks(kmt)',\n",
       " '11_diffExports(kmt)',\n",
       " '11_diffImports(kmt)',\n",
       " '11_diffRefinery intake(kmt)',\n",
       " '11_diffWTI',\n",
       " '11_diffSumClosing stocks(kmt)',\n",
       " '11_diffSumExports(kmt)',\n",
       " '11_diffSumImports(kmt)',\n",
       " '11_diffSumProduction(kmt)',\n",
       " '11_diffSumRefinery intake(kmt)',\n",
       " '12_diffClosing stocks(kmt)',\n",
       " '12_diffExports(kmt)',\n",
       " '12_diffImports(kmt)',\n",
       " '12_diffRefinery intake(kmt)',\n",
       " '12_diffWTI',\n",
       " '12_diffSumClosing stocks(kmt)',\n",
       " '12_diffSumExports(kmt)',\n",
       " '12_diffSumImports(kmt)',\n",
       " '12_diffSumProduction(kmt)',\n",
       " '12_diffSumRefinery intake(kmt)']"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Get the list of features :\n",
    "list(Train_Input)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Statistics of the missing data :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ID                                False\n",
       "month                             False\n",
       "country                           False\n",
       "1_diffClosing stocks(kmt)          True\n",
       "1_diffExports(kmt)                False\n",
       "1_diffImports(kmt)                 True\n",
       "1_diffRefinery intake(kmt)        False\n",
       "1_diffWTI                         False\n",
       "1_diffSumClosing stocks(kmt)      False\n",
       "1_diffSumExports(kmt)             False\n",
       "1_diffSumImports(kmt)             False\n",
       "1_diffSumProduction(kmt)          False\n",
       "1_diffSumRefinery intake(kmt)     False\n",
       "2_diffClosing stocks(kmt)          True\n",
       "2_diffExports(kmt)                False\n",
       "2_diffImports(kmt)                 True\n",
       "2_diffRefinery intake(kmt)        False\n",
       "2_diffWTI                         False\n",
       "2_diffSumClosing stocks(kmt)      False\n",
       "2_diffSumExports(kmt)             False\n",
       "2_diffSumImports(kmt)             False\n",
       "2_diffSumProduction(kmt)          False\n",
       "2_diffSumRefinery intake(kmt)     False\n",
       "3_diffClosing stocks(kmt)          True\n",
       "3_diffExports(kmt)                False\n",
       "3_diffImports(kmt)                 True\n",
       "3_diffRefinery intake(kmt)        False\n",
       "3_diffWTI                         False\n",
       "3_diffSumClosing stocks(kmt)      False\n",
       "3_diffSumExports(kmt)             False\n",
       "                                  ...  \n",
       "10_diffClosing stocks(kmt)         True\n",
       "10_diffExports(kmt)               False\n",
       "10_diffImports(kmt)                True\n",
       "10_diffRefinery intake(kmt)       False\n",
       "10_diffWTI                        False\n",
       "10_diffSumClosing stocks(kmt)     False\n",
       "10_diffSumExports(kmt)            False\n",
       "10_diffSumImports(kmt)            False\n",
       "10_diffSumProduction(kmt)         False\n",
       "10_diffSumRefinery intake(kmt)    False\n",
       "11_diffClosing stocks(kmt)         True\n",
       "11_diffExports(kmt)               False\n",
       "11_diffImports(kmt)                True\n",
       "11_diffRefinery intake(kmt)       False\n",
       "11_diffWTI                        False\n",
       "11_diffSumClosing stocks(kmt)     False\n",
       "11_diffSumExports(kmt)            False\n",
       "11_diffSumImports(kmt)            False\n",
       "11_diffSumProduction(kmt)         False\n",
       "11_diffSumRefinery intake(kmt)    False\n",
       "12_diffClosing stocks(kmt)         True\n",
       "12_diffExports(kmt)               False\n",
       "12_diffImports(kmt)                True\n",
       "12_diffRefinery intake(kmt)       False\n",
       "12_diffWTI                        False\n",
       "12_diffSumClosing stocks(kmt)     False\n",
       "12_diffSumExports(kmt)            False\n",
       "12_diffSumImports(kmt)            False\n",
       "12_diffSumProduction(kmt)         False\n",
       "12_diffSumRefinery intake(kmt)    False\n",
       "dtype: bool"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Check is the dataset Train_Input contains NAN values :\n",
    "Train_Input.isnull().any()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ID        False\n",
       "Target    False\n",
       "dtype: bool"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Check is the dataset Train_Outputs contains NAN values :\n",
    "Train_Outputs.isnull().any()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([False], dtype=bool)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Check is the dataset Test_Inputs contains NAN values :\n",
    "np.unique(Test_Input.isnull().any())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have missing data in the training inputs, no missing data in the test set and training outputs !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of cells in Train_Input with missing data is :  4497\n",
      "Number of rows in Train_Input with missing data is :  294\n"
     ]
    }
   ],
   "source": [
    "#Counting cells with missing data :\n",
    "print( 'Number of cells in Train_Input with missing data is : ', sum(Train_Input.isnull().values.ravel()) )\n",
    "\n",
    "#Counting rows that have missing values somewhere :\n",
    "print( 'Number of rows in Train_Input with missing data is : ', sum([True for idx,row in Train_Input.iterrows() if any(row.isnull())]) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The countries with missing data are : \n",
      "\n",
      "the country  33  has  288 missing values\n",
      "the country  41  has  825 missing values\n",
      "the country  42  has  1524 missing values\n",
      "the country  71  has  1860 missing values\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY8AAAEWCAYAAACe8xtsAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3XmYHWWZ9/HvjxAWA5gAGd6QhYSZgAaVOPQgDsgAiiAq\nqO+A4RVBQNCRQRgYZREFF8QZB31lHNEoyL6JsqggAhJQttiBAAkQCRDMRhJASNgCSe7543laKk0v\nVd19+lTTv891netUPbXd55zuc59nqSpFBGZmZlWs0+wAzMxs4HHyMDOzypw8zMysMicPMzOrzMnD\nzMwqc/IwM7PKnDysU5LOk/SNJh1bkn4q6S+SpvfB/n4o6cu92P5kST/pbRx9SdKnJP2h2XHY4LRu\nswOw8iTNA94ETIiIF3LZp4GDImK3JobWCLsAewJj2l5rb0TEZ3u5/Td7G4PVU/6/+nRE3NTsWAYS\n1zwGniHAMc0OoipJQypushUwry8Sh9VPrlkOiO8fSf6R3YEB8eHZWr4N/Luk4e0XSBovKYp/7JKm\n5dpJWzPH7ZK+K+lZSY9J+sdcPl/SUkmHtNvt5pJulLRC0q2Stirs+y152TOS5kg6oLDsPElnS7pO\n0gvA7h3Eu6Wka/P2cyUdkcsPB34CvFvS85K+2sG2lV5LsQlO0uaSfpW3e0bS79u+yCSdIGlhfr1z\nJL03l58m6aJ27/Mhkv4s6SlJXyoca0NJ5+cmt4ckfVHSgo4+zPwe/Ve7smskHZenT5T0aI7nQUkf\n7WQ/XX72ef6wHM9fJN3Q9lnmL/Lv5vdsuaQHJL2tk+NMk3SGpOl53WskbVpYvpOkO/J7e5+k3dpt\ne7qk24EXga072P9YSb+QtEzS05K+n8vXkXSKpCdynBdIenNetlv791fSPEnvy9OnSboib7NC0mxJ\nLXnZhcA44Jf5b+2LhffycEl/Bn4n6deSjm53jPs7+zwGhYjwY4A8gHnA+4BfAN/IZZ8GpuXp8UAA\n6xa2mUaqkgN8ClgFHEqqwXwD+DPwP8D6wPuBFcBGef3z8vyuefn3gD/kZcOA+Xlf6wLvBJ4CJhW2\nfQ7YmfQjZYMOXs9twA+ADYDJwDJgj0Ksf+jivejJa2l7z84AfggMzY/3AAK2za9py8L7+bd5+jTg\nonbv84+BDYHtgZXAW/PybwG3AiOAMcD9wIJOXseu+ZjK8yOAlwox7A9smd/DjwMvAKPav0clPvv9\ngLnAW/PndQpwR162FzADGJ7fh7e2HaODeKcBC4G35b+Bnxfel9HA08A+Od498/zIwrZ/BrbLMQxt\nt+8hwH3Ad/O+NwB2ycsOy/FvDWxE+h+4MC/brf37S/5fKXx2L+e4huTP/66O1m33Xl6Q49gQOAC4\nu7DO9vm1rdfs74VmPVzzGJi+AhwtaWQPtn08In4aEauBy4GxwNciYmVE/BZ4Bfi7wvq/jojbImIl\n8CVSbWAs8CFSs9JPI2JVRNxL+iLZv7DtNRFxe0SsiYiXi0HkfewMnBARL0fETFJt4+AGvpY2rwKj\ngK0i4tWI+H2kb4TVpMQzSdLQiJgXEY92cfyvRsRLEXEf6Utv+1x+APDNiPhLRCwAzupiH78nfVG9\nJ8//M3BnRCwCiIifRcSi/B5eDjwC7NjdG9OBzwJnRMRDEbEK+CYwOdc+XgU2Bt5CSmIPRcTiLvZ1\nYUTMitSk+GXgAKVmyYOA6yLiuhzvjUAr6Uu7zXkRMTv/zbzabr87khLlFyLihfx30TYg4BPAdyLi\nsYh4HjgJmKLyTUp/yHGtBi7ktc+qK6flOF4CrgW2kTQxL/skcHlEvFLy+G84Th4DUETMAn4FnNiD\nzZcUpl/K+2tftlFhfn7huM8Dz5D+wbcC3pWbJ56V9CzpH/z/dLRtB7YEnomIFYWyJ0i/Xhv1Wtp8\nm/Qr9re5uevEvO1c4FjSL9Wlki6TtGUXx3+yMP1i4VhbsvZr7/R9yEnrMuDAXPT/gIvblks6WNLM\nwnv8NmDzLmLqzFbA9wr7eYZUyxgdEb8Dvk+qtS2VNFXSJl3sq/h6niDV3jbPx9i/3d/ELqRE3dG2\n7Y0FnsjJrb0t87GKx10X2KKL/RW1/6w2KJF4in/7L5N+oByUmzgPJCWhQcvJY+A6FTiCtb9s2zqX\n31QoK36Z98TYtglJGwGbAotI/1i3RsTwwmOjiPiXwrZdXbJ5EbCppI0LZeNITSINFRErIuL4iNga\n2Bc4rq1vIyIuiYhdSF+EAfxHDw6xmNRc1WZsZytmlwL/nGsB7yLV4MjzPwb+FdgsIoYDs0hf+u11\n99nPBz7T7vPaMCLuAIiIsyJiB2ASsA3whS7iLb6ecaSay1P5GBe2O8awiPhWYf2u/ibmA+M6+VJf\nRPpMisddRfoB8ULxdedaUJVaeWcxtS8/n/QD6b3AixFxZ4VjvOE4eQxQ+Vfy5cDnC2XLSF++B0ka\nIukw4G97eah9JO0iaT3g66S24vmkms82kj4paWh+/IOkt5aMfz5wB3CGpA0kvQM4HLiol/F2S9KH\nJP2dJJH6ZVYDayRtK2kPSeuT2shfAtb04BBXACdJGiFpNOnLv1O5ye8pUrPdDRHxbF40jPQFtizH\nfSip5tHRPrr77H+YY9ou7+vNkvbP0/8g6V2ShpK+iF/u5nUfJGmSpDcBXwOuzM1BFwEflrRXjmGD\n3Jk9pot9FU0nJd5vSRqWt985L7sU+DdJE/KPmG+Smo1WAX8i1SQ+mF/DKaTmx7KW0EHnfXs5WawB\nzmSQ1zrAyWOg+xrpC6boCNKvxqdJHZN39PIYl5BqOc8AO5DatcnNTe8HppB+FT5J+pVe5Z/2QFLn\n5CLgKuDU6J+x9hOBm4DngTuBH0TELaTYv0X6In8S+BtS23pVXwMWAI/n41xJ6lDvyiWkwRCXtBVE\nxIOkL6o7SV9wbwdu72IfnX72EXEV6fO5TNJyUg3mA3nxJqQazl9IzUFPk5r2OnMhaQDCk6RO7c/n\nY8wndcyfTEp483M8pb5ncgL6MKmf6s+k9/DjefG5+bi3kd7Xl4Gj83bPAZ8jJd+FpATY4ei2TpwB\nnJKb2v69m3UvIH0ODf+RU3dtIzzMrEEk/QswJSL+qdmx9JakaaTRVbU6276/SDoYODI3bQ5qrnmY\n9TFJoyTtnM9N2BY4nlSzsgEsN9N9Dpja7FjqwMnDrO+tB/yIdJ7J74BrSOez2AAlaS9SU9wSCk2L\ng5mbrczMrDLXPMzMrLI37AW/Nt988xg/fnyzwzAzG1BmzJjxVER0e57MGzZ5jB8/ntbW1maHYWY2\noEh6ovu13GxlZmY94ORhZmaVOXmYmVllTh5mZlaZk4eZmVXm5GFmZpU5eZiZWWVOHmZmVpmTh5mZ\nVfaGPcPczGygUOHGwgPlWrWueZiZWWVOHmZmVpmTh5mZVebkYWZmlTl5mJlZZU4eZmZWmZOHmZlV\n5uRhZmaVNSx5SDpX0lJJswpll0uamR/zJM3M5eMlvVRY9sPCNjtIekDSXElnScXTaczMrBkaeYb5\necD3gQvaCiLi423Tks4Enius/2hETO5gP2cDRwB3A9cBewPXNyBeMzMrqWE1j4i4DXimo2W59nAA\ncGlX+5A0CtgkIu6KiCAloo/0daxmZlZNs/o83gMsiYhHCmUTcpPVrZLek8tGAwsK6yzIZR2SdKSk\nVkmty5Yt6/uozcwMaF7yOJC1ax2LgXG52eo44BJJm1TdaURMjYiWiGgZOXJkH4VqZmbt9ftVdSWt\nC3wM2KGtLCJWAivz9AxJjwLbAAuBMYXNx+QyMzNrombUPN4HPBwRf22OkjRS0pA8vTUwEXgsIhYD\nyyXtlPtJDgauaULMZmZW0MihupcCdwLbSlog6fC8aAqv7yjfFbg/D929EvhsRLR1tn8O+AkwF3gU\nj7QyM2s6xUC580hFLS0t0dra2uwwzMy6VaebQUmaEREt3a3nM8zNzKwyJw8zM6vMycPMzCpz8jAz\ns8qcPMzMrDInDzMzq8zJw8zMKnPyMDOzypw8zMysMicPMzOrzMnDzMwqc/IwM7PKnDzMzKwyJw8z\nM6vMycPMzCpz8jAzs8qcPMzMrLJuk4ek/SVtnKdPkfQLSX/f+NDMzKyuytQ8vhwRKyTtArwPOAc4\nu7FhmZlZnZVJHqvz8weBqRHxa2C97jaSdK6kpZJmFcpOk7RQ0sz82Kew7CRJcyXNkbRXoXwHSQ/k\nZWdJxbv9mg0u0msPs2YqkzwWSvoR8HHgOknrl9zuPGDvDsq/GxGT8+M6AEmTgCnAdnmbH0gaktc/\nGzgCmJgfHe3TzMz6UZkkcABwA7BXRDwLbAp8obuNIuI24JmScewHXBYRKyPicWAusKOkUcAmEXFX\nRARwAfCRkvs0M7MG6TZ5RMSLwFJgl1y0CnikF8c8WtL9uVlrRC4bDcwvrLMgl43O0+3LOyTpSEmt\nklqXLVvWixDNzKwrZUZbnQqcAJyUi4YCF/XweGcDWwOTgcXAmT3cT4ciYmpEtEREy8iRI/ty12Zm\nVlCm2eqjwL7ACwARsQjYuCcHi4glEbE6ItYAPwZ2zIsWAmMLq47JZQvzdPtyMzNrojLJ45Xc3xAA\nkob19GC5D6PNR4G2kVjXAlMkrS9pAqljfHpELAaWS9opj7I6GLimp8c3M7O+sW6Jda7Io62GSzoC\nOIxUa+iSpEuB3YDNJS0ATgV2kzSZlIjmAZ8BiIjZkq4AHiT1qRwVEW1DhD9HGrm1IXB9fpiZWRMp\nVSq6WUnaE3g/IOCGiLix0YH1VktLS7S2tjY7DLM+VTy/o8S/rg0QdfpcJc2IiJbu1itT8yAni9on\nDDMz6x/dJg9JK8j9HaQzy4cCL0TEJo0MzMzM6qvb5BERfx1ZlTut9wN2amRQZmZWb5UuyR7J1cBe\n3a5sZmZvWGWarT5WmF0HaAFeblhEZmZWe2U6zD9cmF5FGmK7X0OiMTOzAaFMn8eh/RGImZkNHJ0m\nD0n/zWujrF4nIj7fkIjMzKz2uqp5+Aw7MzPrUKfJIyLO789AzMxs4Cgz2mok6ZLsk4AN2sojYo8G\nxmVmZjVW5jyPi4GHgAnAV0mjrf7YwJjMzKzmyiSPzSLiHODViLg1Ig4DXOswMxvEypzn8Wp+Xizp\ng8Ai0n3MzcxskCqTPL4h6c3A8cB/A5sA/9bQqMzMrNbKJI+7I+I54Dlg9wbHY2ZmA0CZPo/bJf1W\n0uGSRjQ8IjMzq71uk0dEbAOcAmwHzJD0K0kHNTwyMzOrrVKXZI+I6RFxHLAj8AzQ7QmEks6VtFTS\nrELZtyU9LOl+SVdJGp7Lx0t6SdLM/PhhYZsdJD0gaa6ks/I9RczMrIm6TR6SNpF0iKTrgTuAxaQk\n0p3zgL3bld0IvC0i3gH8CTipsOzRiJicH58tlJ8NHAFMzI/2+zQzs35WpuZxHzAZ+FpEbBMRJ0TE\njO42iojbSLWUYtlvI2JVnr0LGNPVPiSNAjaJiLsiIoALgI+UiNnMzBqozGirrfMXd187DLi8MD9B\n0kzSqK5TIuL3wGhgQWGdBbnMzMyaqMz9PPo8cUj6EunGUhfnosXAuIh4WtIOwNWStuvBfo8EjgQY\nN25cX4VrZmbtVLqHeV+Q9CngQ8An2hJTRKyMiKfz9AzgUWAbYCFrN22NyWUdioipEdESES0jR45s\n0CswM7N+TR6S9ga+COwbES8WykdKGpKntyZ1jD8WEYuB5ZJ2yqOsDgau6c+Yzczs9cpckv2sDoqf\nA1ojotMvckmXArsBm0taAJxKGl21PnBjHnF7Vx5ZtSvwNUmvAmuAz0ZEW2f750gjtzYErs8PMzNr\nojId5hsAbwF+luf/L/A4sL2k3SPi2I42iogDOyg+p5N1fw78vJNlrcDbSsRpZmb9pEzyeAewc0Ss\nBpB0NvB7YBfggQbGZmZmNVWmz2MEsFFhfhiwaU4mKxsSlZmZ1VqZmsd/AjMlTQNE6p/4pqRhwE0N\njM3MzGqqzHke50i6jtcuSXJyRCzK019oWGRmZlZbZYfqrgMsA/4C/J2kXRsXkpmZ1V2Zobr/AXwc\nmE0aRgsQwG0NjMvMzGqsTJ/HR4BtI8Kd42ZmBpRrtnoMGNroQMzMbOAoU/N4kTTa6mYKQ3Mj4vMN\ni8rMzGqtTPK4Nj/MzMyAckN1u73lrJmZDS6dJg9JV0TEAZIeII2u+usi0m0+3tHw6MzMrJa6qnkc\nk58/1B+BmJnZwNHpaKt8Lw2Ap4D5EfEE6XLq2wOLOtvOzMze+MoM1b0N2EDSaOC3wCdJ99cwM7NB\nqkzyUL7r38eAH0TE/kDl+4ubmdkbR6nkIendwCeAX+eyIY0LyczM6q5M8jiWdPvYqyJidr7H+C2N\nDcvMzOqszHketwK3AkhaB3jKZ5ebmQ1u3dY8JF0iaZN886dZwIOSur2Ph6RzJS2VNKtQtqmkGyU9\nkp9HFJadJGmupDmS9iqU7yDpgbzsLEmq/jLNzKwvlWm2mhQRy0lX170emEAacdWd84C925WdCNwc\nEROBm/M8kiYBU0gd8XsDP5DU1q9yNnAEMDE/2u/TzMz6WZnkMVTSUFLyuDYiXmXtM847FBG3Ac+0\nK94PaLvcyfl5n23ll0XEyoh4HJgL7ChpFLBJRNwVEQFcUNjGzMyapEzy+BEwDxgG3CZpK2B5D4+3\nReHkwyeBLfL0aGB+Yb0FuWx0nm5f3iFJR0pqldS6bNmyHoZoZmbd6TZ5RMRZETE6IvaJ5Alg994e\nONckuq3BVNzn1IhoiYiWkSNH9uWuzcysoKsLIx4UERdJOq6TVb7Tg+MtkTQqIhbnJqmluXwhMLaw\n3phctjBPty83M7Mm6qrmMSw/b9zJoyeuBQ7J04cA1xTKp0haX9IEUsf49NzEtVzSTnmU1cGFbczM\nrEk6rXlExI/y81d7smNJlwK7AZtLWgCcCnwLuELS4cATwAH5GLMlXQE8CKwCjoqI1XlXnyON3NqQ\nNNrr+p7EY2ZmfUep66GLFVJN4GhgPIVkExH7NjSyXmppaYnW1tZmh2HWp4pnOXXzr2sDSJ0+V0kz\nIqKlu/XK3Ib2auAc4JfAmt4GZmZmA1+Z5PFyRJzV8EjMzGzAKJM8vifpVNK9PFa2FUbEPQ2LyszM\naq1M8ng76XIke/Bas1XkeTMzG4TKJI/9ga0j4pVGB2NmZgNDmcuTzAKGNzoQMzMbOMrUPIYDD0v6\nI2v3edR6qK6ZmTVOmeRxasOjMDOzAaXsnQTNzMz+qkyfh5mZ2VqcPMzMrLJOk4ekm/Pzf/RfOGZm\nNhB01ecxStI/AvtKugxQcaHPMDczG7y6Sh5fAb5MugFT+xs/+QxzM7NBrKv7eVwJXCnpyxHx9X6M\nycyyOl2q26yozFDdr0vaF9g1F02LiF81NiwzM6uzbkdbSToDOIZ0l78HgWMkfbPRgZmZWX2VOcP8\ng8DkiFgDIOl84F7g5EYGZmZm9VX2PI/ihRHf3IhAzMxs4ChT8zgDuFfSLaThursCJ/b0gJK2BS4v\nFG1NGtk1HDgCWJbLT46I6/I2JwGHA6uBz0fEDT09vpmZ9V6ZDvNLJU0D/iEXnRART/b0gBExB5gM\nIGkIsBC4CjgU+G5E/FdxfUmTgCnAdsCWwE2StomI1T2NwczMeqdMzYOIWAxc24Djvxd4NCKekNTZ\nOvsBl0XESuBxSXOBHYE7GxCPmZmV0OxrW00BLi3MHy3pfknnShqRy0YD8wvrLMhlryPpSEmtklqX\nLVvW0SpmZtYHmpY8JK0H7Av8LBedTer/mAwsBs6sus+ImBoRLRHRMnLkyD6L1czM1tZl8pA0RNLD\nDTr2B4B7ImIJQEQsiYjVeUjwj0lNU5D6RMYWthuTy8zMrEm6TB65U3qOpHENOPaBFJqsJI0qLPso\n6d7pkPpapkhaX9IEYCIwvQHxmJlZSWU6zEcAsyVNB15oK+zNPcwlDQP2BD5TKP5PSZNJF12c17Ys\nImZLuoJ0dvsq4CiPtDIza64yyePLfX3QiHgB2Kxd2Se7WP904PS+jsPMzHqm1D3MJW0FTIyImyS9\nCRjS+NDMzKyuylwY8QjgSuBHuWg0cHUjgzIzs3orM1T3KGBnYDlARDwC/E0jgzIzs3orkzxWRsQr\nbTOS1iV1apuZ2SBVJnncKulkYENJe5JO6vtlY8MyM7M6K5M8TiRd6fYB0vDZ64BTGhmUmZnVW5nR\nVmvyDaDuJjVXzYnw3ZTNzAazbpOHpA8CPwQeJd3PY4Kkz0TE9Y0OzszM6qnMSYJnArtHxFwASX8L\n/Bpw8jAzG6TK9HmsaEsc2WPAigbFY2ZmA0CnNQ9JH8uTrZKuA64g9XnsD/yxH2IzM7Oa6qrZ6sOF\n6SXAP+XpZcCGDYvIzMxqr9PkERGH9mcgZmY2cJQZbTUBOBoYX1y/N5dkNzOzga3MaKurgXNIZ5Wv\naWw4ZmY2EJRJHi9HxFkNj8TMzAaMMsnje5JOBX4LrGwrjIh7GhaVmZnVWpnk8Xbgk8AevNZsFXne\nzMwGoTLJY39g6+Jl2c3MbHArc4b5LGB4Xx5U0jxJD0iaKak1l20q6UZJj+TnEYX1T5I0V9IcSXv1\nZSxmZlZdmZrHcOBhSX9k7T6P3g7V3T0inirMnwjcHBHfknRinj9B0iRgCrAdsCVwk6RtImJ1L49v\nZmY9VCZ5nNrwKJL9gN3y9PnANOCEXH5ZRKwEHpc0F9gRuLOf4jIzs3bK3M/j1gYcN0g1iNXAjyJi\nKrBFRCzOy58EtsjTo4G7CtsuyGWvI+lI4EiAcePGNSBsMzODcmeYr+C1e5avBwwFXoiITXpx3F0i\nYqGkvwFulPRwcWFEhKTKN5zKSWgqQEtLi29YZWbWIGVqHhu3TUsSqRlpp94cNCIW5uelkq4iNUMt\nkTQqIhZLGgUszasvBMYWNh+Ty8zMrEnKjLb6q0iuBno84knSMEkbt00D7yeN6LoWOCSvdghwTZ6+\nFpgiaf18na2JwPSeHt/MzHqvTLPVxwqz6wAtwMu9OOYWwFWpEsO6wCUR8Zs8musKSYcDTwAHAETE\nbElXAA8Cq4CjPNLKzKy5yoy2Kt7XYxUwj9R01SMR8RiwfQflTwPv7WSb04HTe3pMMzPrW2X6PHxf\nDzMzW0tXt6H9ShfbRUR8vQHxmJnZANBVzeOFDsqGAYcDmwFOHmZmg1RXt6E9s206j446BjgUuAw4\ns7PtzMzsja/LPg9JmwLHAZ8gXTLk7yPiL/0RmJmZ1VdXfR7fBj5GOmP77RHxfL9FZWZmtdbVSYLH\nk65iewqwSNLy/FghaXn/hGdmZnXUVZ9HpbPPzcxs8HCCMDOzypw8zMysMicPMzOrzMnDzMwqc/Iw\nM7PKnDzMzKwyJw8zM6vMycPMzCpz8jAzs8qcPMzMrLIyt6E1G5Sk16YjmheHWR31e81D0lhJt0h6\nUNJsScfk8tMkLZQ0Mz/2KWxzkqS5kuZI2qu/YzYzs7U1o+axCjg+Iu7JN5maIenGvOy7EfFfxZUl\nTQKmANuRrvJ7k6RtImJ1v0ZtZmZ/1e81j4hYHBH35OkVwEPA6C422Q+4LCJWRsTjwFxgx8ZHamZm\nnWlqh7mk8cA7gbtz0dGS7pd0rqQRuWw0ML+w2QI6STaSjpTUKql12bJlDYrazMyaljwkbQT8HDg2\nIpYDZwNbA5OBxfTgPukRMTUiWiKiZeTIkX0ar5mZvaYpyUPSUFLiuDgifgEQEUsiYnVErAF+zGtN\nUwuBsYXNx+QyMzNrkmaMthJwDvBQRHynUD6qsNpHgVl5+lpgiqT1JU0AJgLT+yteMzN7vWaMttoZ\n+CTwgKSZuexk4EBJk4EA5gGfAYiI2ZKuAB4kjdQ6yiOtzMyaq9+TR0T8AVAHi67rYpvTgdMbFpSZ\nmVXiy5OYmVllTh5mZlaZk4eZmVXm5GFmZpU5eZiZWWVOHmZmVpmTh5mZVebkYWZmlTl5mJlZZU4e\nZmZWmZOHmZlV5uRhZmaVOXmYmVllTh5mZlaZk4eZmVXm5GFmZpU5eZiZWWVOHmZmVpmTh5mZVTZg\nkoekvSXNkTRX0onNjsfMbDAbEMlD0hDgf4APAJOAAyVNam5UZmaD14BIHsCOwNyIeCwiXgEuA/Zr\nckxmZoPWus0OoKTRwPzC/ALgXe1XknQkcGSefV7SnArH2Bx4qscR9g/H2Dcqxyg1KJKOdRhf+xj6\nOab23pCfcz8r9Tk3wVZlVhooyaOUiJgKTO3JtpJaI6Klj0PqU46xb9Q9xrrHB46xL9Q9vu4MlGar\nhcDYwvyYXGZmZk0wUJLHH4GJkiZIWg+YAlzb5JjMzAatAdFsFRGrJP0rcAMwBDg3Imb38WF61NzV\nzxxj36h7jHWPDxxjX6h7fF1SRDQ7BjMzG2AGSrOVmZnViJOHmZlV5uRBPS99IulcSUslzSqUbSrp\nRkmP5OcRTYxvrKRbJD0oabakY2oY4waSpku6L8f41brFmOMZIuleSb+qaXzzJD0gaaak1prGOFzS\nlZIelvSQpHfXKUZJ2+b3r+2xXNKxdYqxqkGfPGp86ZPzgL3blZ0I3BwRE4Gb83yzrAKOj4hJwE7A\nUfl9q1OMK4E9ImJ7YDKwt6SdahYjwDHAQ4X5usUHsHtETC6cl1C3GL8H/CYi3gJsT3o/axNjRMzJ\n799kYAfgReCqOsVYWUQM6gfwbuCGwvxJwEnNjivHMh6YVZifA4zK06OAOc2OsRDbNcCedY0ReBNw\nD+nKBLWJkXTO0s3AHsCv6vg5A/OAzduV1SZG4M3A4+QBQHWMsV1c7wdur3OMZR6DvuZBx5c+Gd2k\nWLqzRUQsztNPAls0M5g2ksYD7wTupmYx5iahmcBS4MaIqFuM/x/4IrCmUFan+AACuEnSjHwJIKhX\njBOAZcBPc/PfTyQNo14xFk0BLs3TdY2xW04eA1SknypNH2ctaSPg58CxEbG8uKwOMUbE6khNBWOA\nHSW9rd3ypsUo6UPA0oiY0dk6dXgPgV3ye/gBUvPkrsWFNYhxXeDvgbMj4p3AC7Rr/qlBjADkk5z3\nBX7WflldYizLyWNgXfpkiaRRAPl5aTODkTSUlDgujohf5OJaxdgmIp4FbiH1I9Ulxp2BfSXNI10p\neg9JF9WWFQ4RAAADX0lEQVQoPgAiYmF+Xkpqp9+ResW4AFiQa5UAV5KSSZ1ibPMB4J6IWJLn6xhj\nKU4eA+vSJ9cCh+TpQ0j9DE0hScA5wEMR8Z3CojrFOFLS8Dy9IalP5mFqEmNEnBQRYyJiPOnv7ncR\ncVBd4gOQNEzSxm3TpPb6WdQoxoh4Epgvadtc9F7gQWoUY8GBvNZkBfWMsZxmd7rU4QHsA/wJeBT4\nUrPjyTFdCiwGXiX9sjoc2IzUufoIcBOwaRPj24VUxb4fmJkf+9QsxncA9+YYZwFfyeW1ibEQ6268\n1mFem/iArYH78mN22/9HnWLM8UwGWvNnfTUwooYxDgOeBt5cKKtVjFUevjyJmZlV5mYrMzOrzMnD\nzMwqc/IwM7PKnDzMzKwyJw8zM6vMycOsn+Wrqb6pl/v4SUcX8JT0KUnf782+zcpw8jDrf8eSLtT4\nOvkqz92KiE9HxIN9GpVZBU4eZu1IOljS/fk+IBfmsvGSfpfLb5Y0LpefJ+mfC9s+n593kzStcI+J\ni5V8HtgSuEXSLW3bSDpT0n3AlyRdXdjfnpKu6iDGaZJa8vShkv4kaTrpkidmDbduswMwqxNJ2wGn\nAP8YEU9J2jQv+m/g/Ig4X9JhwFnAR7rZ3TuB7YBFwO3AzhFxlqTjSPfHeCqvNwy4OyKOz5d9eUjS\nyIhYBhwKnNtFvKOAr5LuEfEc6fpd91Z/5WbVuOZhtrY9gJ+1fbFHxDO5/N3AJXn6QtLlWbozPSIW\nRMQa0uVbxney3mrSBSaJdMmHC4GD8nW53g1c38Ux3gVMi4hlEfEKcHmJuMx6zTUPs95ZRf4RJmkd\nYL3CspWF6dV0/v/2ckSsLsz/FPgl8DIpka3qu3DN+oZrHmZr+x2wv6TNIN2rO5ffQbryLcAngN/n\n6XmkJiNI92kYWuIYK4CNO1sYEYtITV2nkBJJV+4G/knSZvkS+fuXOL5Zr7nmYVYQEbMlnQ7cKmk1\nqf/gU8DRpDvVfYF017pD8yY/Bq7Jnd2/Id2IqDtTgd9IWhQRu3eyzsXAyIh4qJPlbfEulnQacCfw\nLKl5zKzhfFVdsxrK52rcGxHnNDsWs444eZjVjKQZpBrMnhGxsrv1zZrBycPMzCpzh7mZmVXm5GFm\nZpU5eZiZWWVOHmZmVpmTh5mZVfa/duJQTHl1/CkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1160a2e80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## missing values per country\n",
    "missing_data_per_country = []\n",
    "for i in np.arange(1, len(np.unique(Train_Input.iloc[:, 2]))):\n",
    "    missing_data_per_country.append([i, sum( Train_Input[ Train_Input['country'] == i].isnull().values.ravel()) ])\n",
    "missing_data_per_country = np.array(missing_data_per_country)\n",
    "\n",
    "#_Print the countries with missing values :\n",
    "print('The countries with missing data are : \\n')\n",
    "for i in range(0, len(missing_data_per_country)):\n",
    "    if missing_data_per_country[i,1] != 0 :\n",
    "        print('the country ', missing_data_per_country[i,0], ' has ', missing_data_per_country[i,1], 'missing values')\n",
    "print('\\n')\n",
    "\n",
    "#_Diragram showing the number of missing values per country\n",
    "plt.bar(missing_data_per_country[:,0], missing_data_per_country[:,1], color=\"blue\")\n",
    "plt.ylabel('Number of missing values')\n",
    "plt.xlabel('country id')\n",
    "plt.title('Number of missing values per country')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3gAAAJcCAYAAACrJAbaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEvDazFRmwAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11a9c12e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Get the distribution of features with missing data for the selected countries\n",
    "\n",
    "countries_missing = [33, 41, 42, 71] #countries with mising data\n",
    "Missing_data_per_features ={} #dictionnary in order to get the statistics about missing data/feature\n",
    "Features_list = list(Train_Input) #list of features\n",
    "\n",
    "#Fill the dictionnary with the number of missing data per feature :\n",
    "for country in countries_missing : #loop over the features\n",
    "    \n",
    "    selected_country = Train_Input['country']==country #coundition to select data for the following country\n",
    "    Missing_data_per_features[country] = [[Feature, int(sum(Train_Input[selected_country][Feature].isnull().values.ravel()))] \\\n",
    "                                          for Feature in Features_list]\n",
    "    Missing_data_per_features[country] = np.array(Missing_data_per_features[country])\n",
    "\n",
    "\n",
    "#plot the number of missing data per feature for each country :\n",
    "x = range(len(Features_list))\n",
    "# row and column sharing\n",
    "f, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2, sharex='col', sharey='row', figsize=(15, 10))\n",
    "\n",
    "ax1.bar(x, list(map(int, Missing_data_per_features[33][:,1])), color=\"blue\", align='center')\n",
    "ax1.set_title('Number of missing values per feature for the country 33')\n",
    "ax1.set_xlabel('Features')\n",
    "ax1.set_ylabel('Number of missing data')\n",
    "\n",
    "ax2.bar(x, list(map(int, Missing_data_per_features[41][:,1])), color=\"blue\", align='center')\n",
    "ax2.set_title('Number of missing values per feature for the country 41')\n",
    "ax2.set_xlabel('Features')\n",
    "ax2.set_ylabel('Number of missing data')\n",
    "\n",
    "ax3.bar(x, list(map(int, Missing_data_per_features[42][:,1])), color=\"blue\", align='center')\n",
    "ax3.set_title('Number of missing values per feature for the country 42')\n",
    "ax3.set_xlabel('Features')\n",
    "ax3.set_ylabel('Number of missing data')\n",
    "\n",
    "ax4.bar(x, list(map(int, Missing_data_per_features[71][:,1])), color=\"blue\", align='center')\n",
    "ax4.set_title('Number of missing values per feature for the country 71')\n",
    "ax4.set_xlabel('Features')\n",
    "ax4.set_ylabel('Number of missing data')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Replace the missing values :\n",
    "\n",
    "In order to replace the missing data we will follow the strategy :\n",
    "\n",
    "    1. If the data missing os highly correlated to the country : closing_stocks, exports/imports, refinery intake, WTI. we will replace using the mean of the this features in the same country for the selected month.\n",
    "    \n",
    "    2. If the data missing depends on all the countries i.e starting with sum we will fill using a mean of this feature over all the countries for the corresponding month.\n",
    "\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def Fill_Missing_values(Train_Input, country_list):\n",
    "    \n",
    "    for country in country_list :\n",
    "    \n",
    "        Country_data = Train_Input[Train_Input['country']==country]\n",
    "        indexes = list(Country_data.index) #row indexes of Country_data\n",
    "        #Get indexes of missing values :\n",
    "        missing_x, missing_y = np.where(pd.isnull(Country_data))\n",
    "        \n",
    "        for i in range(len(missing_x)):\n",
    "            x,y = missing_x[i], missing_y[i]\n",
    "            #if the feature depends only on the country :\n",
    "            if \"Sum\" not in list(Country_data)[y] :\n",
    "                # First we get the subset of elements with the same month of the missing data point:\n",
    "                same_month_condition = Country_data['month'] == Country_data.iloc[x,1] \n",
    "                same_month_data = Country_data[same_month_condition]   \n",
    "                # Fill in with the mean value (take care to not consider the missing values) :\n",
    "                Train_Input.at[indexes[x], list(Country_data)[y]] = np.mean([ same_month_data.iloc[:,y].dropna() ])\n",
    "            \n",
    "                #if the feature depends on all the countries i.e sum is in the column name :\n",
    "            if \"Sum\" in list(Country_data)[y] :    \n",
    "                # First get the subset of elements for the same month from all the dataset:\n",
    "                same_month_condition = Train_Input[Train_Input['month']] == Country_data.iloc[x,1]                          \n",
    "                same_month_data =  Train_Input[same_month_condition]\n",
    "                # Fill in with the mean value over all the dataset\n",
    "                Train_Input.at[indexes[x], list(Country_data)[y]] = np.mean([ same_month_data.iloc[:,y].dropna() ]) \n",
    "        \n",
    "        print('Fill_Missing_values performed for the country', country, ':) \\n')\n",
    "    return"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fill_Missing_values performed for the country 33 :) \n",
      "\n",
      "Fill_Missing_values performed for the country 41 :) \n",
      "\n",
      "Fill_Missing_values performed for the country 42 :) \n",
      "\n",
      "Fill_Missing_values performed for the country 71 :) \n",
      "\n",
      "Security check : \n",
      "\n",
      "the number of missing data in  33  is : 0\n",
      "the number of missing data in  41  is : 0\n",
      "the number of missing data in  42  is : 0\n",
      "the number of missing data in  71  is : 0\n",
      "Total number of missing data in the dataset is :  0\n"
     ]
    }
   ],
   "source": [
    "# Fill the missing values in the dataset :\n",
    "country_list = [33, 41, 42, 71]\n",
    "Fill_Missing_values(Train_Input, country_list)\n",
    "# Secutiry check : see if the data has been replaced :\n",
    "print('Security check : \\n')\n",
    "for country in country_list :\n",
    "    print('the number of missing data in ', country, ' is :', sum( Train_Input[ Train_Input['country'] == country].isnull().values.ravel()) )\n",
    "    \n",
    "print('Total number of missing data in the dataset is : ', sum( Train_Input.isnull().values.ravel()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## lV. Visualization of the data \n",
    "\n",
    "We will visualize the data per country/month using two different algorithms :\n",
    "\n",
    "    1. PCA : a linear transformation that maximize the variance.\n",
    "\n",
    "    2. t-SNE : a non linear transformation that works well with high dimensional data, it preserves the distances between the data in the new space.\n",
    "\n",
    "Multiple dimentionnality reduction methods are available, for an exhaustif discussion please se [Colah's blog](http://colah.github.io/posts/2014-10-Visualizing-MNIST/).\n",
    "\n",
    "Note : the algorithm should ignore the country/id/month data !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "'''\n",
    " ***Inputs :\n",
    "    - Train_Input : All the dataset\n",
    "    - show_by_months : bool (true or false) to determine if we want to see the distribution per month\n",
    "    - selected_months : list of months that we want to consider if select_per_month==True\n",
    "    - show_by_countries : bool (true or false) to determine if we want to see the distribution per country\n",
    "    - selected_countries : list of countries that we want to consider if select_per_countries==True\n",
    "    - figure_size : size of the figure = (5,5) by default\n",
    "    \n",
    "    *** Note : if show_per_month is always different than show_by_countries : \n",
    "'''\n",
    "\n",
    "def plot_pca_distribution_per_country(Train_Input, \n",
    "                                      show_by_months, selected_months,\n",
    "                                      show_by_countries, selected_countries,\n",
    "                                      figure_size=(10,10)):   \n",
    "    ##Safety check :\n",
    "    #Check if show_per_month is different than show_by_countries\n",
    "    if (show_by_months and show_by_countries)==True:\n",
    "        print('please select only one feature to display : months or country, not both')\n",
    "        return \n",
    "    # if no feature is selected\n",
    "    if (show_by_months or show_by_countries)==False:\n",
    "        print('please select one feature to display : months or country, not both')\n",
    "        return \n",
    "    \n",
    "    print('The selected Months are : ', selected_months)\n",
    "    print('The selected Countries are : ', selected_countries)\n",
    "    # get the data for the selected country\n",
    "    subset_selected = Train_Input['month'].isin(selected_months) &  Train_Input['country'].isin(selected_countries)\n",
    "    Data = Train_Input[subset_selected]\n",
    "       \n",
    "    #perform PCA on this data :\n",
    "    # Do not consider the features id, country and month\n",
    "    pca = PCA(n_components=2)\n",
    "    reduction = pca.fit_transform(Data.iloc[:,3:])\n",
    "\n",
    "    #Get the reduction with the panda dataframe :\n",
    "    Reduction_dataframe = Data.iloc[:,:3]\n",
    "    Reduction_dataframe['x_transformation'] = reduction[:,0]\n",
    "    Reduction_dataframe['y_transformation'] = reduction[:,1]\n",
    "\n",
    "    # plot in 2d space :\n",
    "    fig = plt.figure(figsize=figure_size)\n",
    "    ax = plt.subplot(111)\n",
    "\n",
    "    if show_by_countries :\n",
    "        for country in selected_countries:\n",
    "            cond = Reduction_dataframe['country'] == country\n",
    "            ax.plot(Reduction_dataframe[cond].iloc[:,3], Reduction_dataframe[cond].iloc[:,4], \\\n",
    "                     linestyle='none', marker='o', markersize=4, label='country %i'%country)\n",
    "        \n",
    "        # Shrink current axis's height by 10% on the bottom\n",
    "        box = ax.get_position()\n",
    "        ax.set_position([box.x0, box.y0 + box.height * 0.1,\n",
    "                         box.width, box.height * 0.9])\n",
    "\n",
    "        # Put a legend below current axis\n",
    "        ax.legend(loc='upper center', bbox_to_anchor=(0.5, -0.05),\n",
    "        fancybox=True, shadow=True, ncol=5)\n",
    "        plt.title('Distribution of data for the selected countries-months, separated by country using PCA')\n",
    "        plt.show()\n",
    "        return\n",
    "    \n",
    "    if show_by_months :\n",
    "        for month in selected_months:\n",
    "            cond = Reduction_dataframe['month'] == month\n",
    "            ax.plot(Reduction_dataframe[cond].iloc[:,3], Reduction_dataframe[cond].iloc[:,4], \\\n",
    "                     linestyle='none', marker='o', markersize=2, label='month %i'%month)\n",
    "        #'markersize=1,'   \n",
    "        # Shrink current axis's height by 10% on the bottom\n",
    "        box = ax.get_position()\n",
    "        ax.set_position([box.x0, box.y0 + box.height * 0.1,\n",
    "                         box.width, box.height * 0.9])\n",
    "        # Put a legend below current axis\n",
    "        ax.legend(loc='upper center', bbox_to_anchor=(0.5, -0.05),\n",
    "        fancybox=True, shadow=True, ncol=5)                 \n",
    "        plt.title('Distribution of data for the selected countries-months, separated by month using PCA')\n",
    "        plt.show()\n",
    "        return"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Distribution of data by month "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The selected Months are :  [ 1  2  3  4  5  6  7  8  9 10 11 12]\n",
      "The selected Countries are :  [ 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25\n",
      " 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50\n",
      " 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAm8AAAJVCAYAAACI1xJDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115b35b70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_by_months = True\n",
    "selected_months = np.arange(1,13)\n",
    "show_by_countries = False\n",
    "selected_countries = np.arange(1,number_of_countries)\n",
    "plot_pca_distribution_per_country(Train_Input, \n",
    "                                      show_by_months, selected_months,\n",
    "                                      show_by_countries, selected_countries)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Distribution of data by country "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The selected Months are :  [ 1  2  3  4  5  6  7  8  9 10 11 12]\n",
      "The selected Countries are :  [1 2 3 4]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAm8AAAI3CAYAAADN67wSAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11a588860>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_by_months = False\n",
    "selected_months = np.arange(1,13)\n",
    "show_by_countries = True\n",
    "selected_countries = np.arange(1,5)\n",
    "plot_pca_distribution_per_country(Train_Input, \n",
    "                                      show_by_months, selected_months,\n",
    "                                      show_by_countries, selected_countries)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "One can see that the separation between months is almost perfect using a PCA, we can see the different classes. When we try to do the same with countries we can see that it's very difficult to see any separation between the points. We explain this behavior by the fact that our data is highly dependent on the weather and then on the months. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# V. Classification"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Split the data :\n",
    "\n",
    "In order to evaluate properly our model, we need to split the dataset into (X_train, Y_train) and (X_validation, Y_validation).\n",
    "\n",
    "-> (X_train, Y_train) is used as a training dataset for our model.\n",
    "\n",
    "-> (X_validation, Y_validation) is used to evaluate the trained model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train shape is  (8127, 122)\n",
      "Y_train shape is  (8127,)\n",
      "X_validation shape is  (2032, 122)\n",
      "Y_validation shape is  (2032,)\n"
     ]
    }
   ],
   "source": [
    "## Variables for our study :\n",
    "validation_size = 0.20\n",
    "seed = 7\n",
    "num_folds = 10\n",
    "\n",
    "## split the data :\n",
    "# Note : we need to remove the Id-cells\n",
    "X_train, X_validation, Y_train, Y_validation = train_test_split(Train_Input.iloc[:,1:], Train_Outputs.iloc[:,1], \\\n",
    "                                                                test_size=validation_size, random_state=seed)\n",
    "\n",
    "# Get an idea about the sizes :\n",
    "print('X_train shape is ', X_train.shape)\n",
    "print('Y_train shape is ', Y_train.shape)\n",
    "print('X_validation shape is ', X_validation.shape)\n",
    "print('Y_validation shape is ', Y_validation.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Algorithms :\n",
    "\n",
    "We will evaluate the performance of different algorithms, and compare them in order to choose the best one."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## List of classification agorithms to try first :\n",
    "pipelines = []\n",
    "\n",
    "# SVM algorithm with different kernels : rbf, linear, polynomial and sigmoid :\n",
    "pipelines.append(('ScaledSVM_rbf', Pipeline([('Scaler', StandardScaler()),('SVM_rbf', svm.SVC(kernel='rbf'))])))\n",
    "pipelines.append(('ScaledSVM_linear', Pipeline([('Scaler', StandardScaler()),('SVM_linear', svm.SVC(kernel='linear'))])))\n",
    "pipelines.append(('ScaledSVM_sigmoid', Pipeline([('Scaler', StandardScaler()),('SVM_sigmoid', svm.SVC(kernel='sigmoid'))])))\n",
    "# Random forest :\n",
    "pipelines.append(('ScaledRF', Pipeline([ ('Scaler', StandardScaler()),('RF', RandomForestClassifier()) ])))\n",
    "# ExtraTrees classifier :\n",
    "pipelines.append(('ScaledET', Pipeline([('Scaler', StandardScaler()),('ET', ExtraTreesClassifier())])))\n",
    "#   radient Boosting classifier :                                  \n",
    "pipelines.append(('ScaledGBM', Pipeline([('Scaler', StandardScaler()),('GBM', GradientBoostingClassifier())])))\n",
    "# K_NN algorithm :\n",
    "pipelines.append(('ScaledKNN', Pipeline([('Scaler', StandardScaler()),('KNN', KNeighborsClassifier(n_neighbors=100))])))\n",
    "# XGboost\n",
    "pipelines.append(('ScaledXGB', Pipeline([('Scaler', StandardScaler()),('XGB', XGBClassifier())])))\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In order to evaluate the algorithms we use cross-validation (k-fold) the number of folds is 10 :\n",
    "\n",
    "Note : we introduce the variable Compare_algorithms=False, if the user wants to see the results of all the algorithms please set his value to True."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 4 µs, sys: 1e+03 ns, total: 5 µs\n",
      "Wall time: 11 µs\n",
      "ScaledSVM_rbf: 0.703395 (0.013770)\n",
      "ScaledSVM_linear: 0.693446 (0.011666)\n",
      "ScaledSVM_sigmoid: 0.602173 (0.023997)\n",
      "ScaledRF: 0.732562 (0.013532)\n",
      "ScaledET: 0.709366 (0.012050)\n",
      "ScaledGBM: 0.781627 (0.014187)\n",
      "ScaledKNN: 0.706051 (0.017993)\n",
      "ScaledXGB: 0.784630 (0.015043)\n"
     ]
    }
   ],
   "source": [
    "Compare_algorithms = True\n",
    "\n",
    "if Compare_algorithms :\n",
    "    %%time\n",
    "    results = []\n",
    "    names = []\n",
    "    for name, model in pipelines:\n",
    "\n",
    "        kfold = KFold(n_splits=num_folds, random_state=seed)\n",
    "        cv_results = cross_val_score(model, X_train, Y_train, cv=kfold, scoring = 'roc_auc')\n",
    "        results.append(cv_results)\n",
    "        names.append(name)\n",
    "\n",
    "        print( \"%s: %f (%f)\" % (name, cv_results.mean(), cv_results.std()) )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Campare the algorithms :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3QAAAHPCAYAAAAbApgkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11fbfd2b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if Compare_algorithms :\n",
    "    fig = plt.figure()\n",
    "    fig.suptitle('Scaled Algorithm Comparison')\n",
    "    ax = fig.add_subplot(111)\n",
    "    plt.boxplot(results)\n",
    "    ax.set_xticklabels(names)\n",
    "    fig.set_size_inches(15,7)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Improve the accuracy - Finetunning the parameters :\n",
    "\n",
    "\n",
    "Build a function using grid search in order to find the best parameters specified in 'params' for a given model 'my_model'."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "'''\n",
    "Inputs :\n",
    "    - my_model : the model to use\n",
    "    - params : the parameters to finetune\n",
    "    \n",
    "Outputs :\n",
    "    The performance of each parameters, and prints the best parameters to use.\n",
    "'''\n",
    "def tune_parameters(my_model, params, X_train, Y_train, seed, split_number=5, print_=True):\n",
    "    scaler = StandardScaler().fit(X_train)\n",
    "    rescaledX = scaler.transform(X_train)\n",
    "    kfold = KFold(n_splits=split_number, random_state=seed)\n",
    "    grid = GridSearchCV(estimator=my_model, param_grid=params, scoring='roc_auc', cv=kfold)\n",
    "    grid_result = grid.fit(rescaledX, Y_train)\n",
    "    print(\"Best: %f using %s\" % (grid_result.best_score_, grid_result.best_params_))\n",
    "    means = grid_result.cv_results_['mean_test_score']\n",
    "    stds = grid_result.cv_results_['std_test_score']\n",
    "    params = grid_result.cv_results_['params']\n",
    "    if print_ :\n",
    "        for mean, stdev, param in zip(means, stds, params):\n",
    "            print(\"%f (%f) with: %r\" % (mean, stdev, param))\n",
    "        \n",
    "    return grid_result.best_score_, grid_result.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finetuning all the parameters at once took us a long time on the local machine, this step needs strong CPUs on the machine.\n",
    "\n",
    "In our code we introduce the choice between finetuning all the parameters at once by letting Strong_CPU=True otherwise we will finetune each parameters in a single run of the function tune_parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "'''\n",
    "Useful function for our greedy parameter search\n",
    "\n",
    "Goal : Transforms the values of a dictionary to a list-type\n",
    "'''\n",
    "def transform_values_into_list(Mydict):\n",
    "    \n",
    "    for key in Mydict :\n",
    "        if isinstance(Mydict[key], (list))==False  :            \n",
    "            Mydict[key] = [Mydict[key]]\n",
    "    return Mydict        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Set Tune_param to True if you want to perform a gridsearch :\n",
    "Tune_param = False\n",
    "#Set Strong_CPU to true if you want to perform a grid search for all the parameters at once otherwise \n",
    "#it will performs a greedy search\n",
    "Strong_CPU = False\n",
    "\n",
    "if Tune_param :\n",
    "    \n",
    "    my_model =  XGBClassifier()\n",
    "\n",
    "    if Strong_CPU :\n",
    "        param_grid = {\n",
    "                    'max_depth':[8, 9, 10, 11],\n",
    "                    'min_child_weight':np.arange(2,4),\n",
    "                    'gamma':[i/10.0 for i in range(0,4)],\n",
    "                    }\n",
    "        best_accuracy, Best_param = tune_parameters(my_model, param_grid, X_train, Y_train, seed) \n",
    "\n",
    "\n",
    "    if not Strong_CPU :\n",
    "            # Check the best param for 'max_depth' :\n",
    "            best_score, greedy_param = tune_parameters(my_model, {'max_depth':[8, 9, 10, 11]}, X_train, Y_train, seed, print_=False) \n",
    "            greedy_param = transform_values_into_list(greedy_param)\n",
    "            print('We get the best score with max_depth :', greedy_param, 'the accuracy is :', best_score)\n",
    "\n",
    "            # Check the best param for 'min_child_weight' :\n",
    "            greedy_param.update({'min_child_weight':np.arange(2,4)})\n",
    "            best_score, greedy_param = tune_parameters(my_model, greedy_param, X_train, Y_train, seed, print_=False) \n",
    "            greedy_param = transform_values_into_list(greedy_param)\n",
    "            print('We get the best score with max_depth :', greedy_param, 'the accuracy is :', best_score)\n",
    "\n",
    "            # Check the best param for 'gamma' :\n",
    "            greedy_param.update({'gamma':[i/10.0 for i in range(0,4)]})\n",
    "            best_score, greedy_param = tune_parameters(my_model, greedy_param, X_train, Y_train, seed, print_=False) \n",
    "            greedy_param = transform_values_into_list(greedy_param)\n",
    "            print('We get the best score with max_depth :', greedy_param, 'the accuracy is :', best_score)\n",
    "        \n",
    "            Best_param = greedy_param"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Get the Finale model :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Best_param = {'gamma': 0.0, 'max_depth': 10, 'min_child_weight': 2}\n",
    "Optimized_Model = Pipeline([('Scaler', StandardScaler()),('XGB', XGBClassifier(gamma=0.0, max_depth=10, min_child_weight=2))])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Train the model on the complete dataset :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Pipeline(steps=[('Scaler', StandardScaler(copy=True, with_mean=True, with_std=True)), ('XGB', XGBClassifier(base_score=0.5, colsample_bylevel=1, colsample_bytree=1,\n",
       "       gamma=0.0, learning_rate=0.1, max_delta_step=0, max_depth=10,\n",
       "       min_child_weight=2, missing=None, n_estimators=100, nthread=-1,\n",
       "       objective='binary:logistic', reg_alpha=0, reg_lambda=1,\n",
       "       scale_pos_weight=1, seed=0, silent=True, subsample=1))])"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Optimized_Model.fit(X_train, Y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Results on the validation set :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "the accuracy is :  0.765255905512\n",
      "the roc-auc score is : 0.714942149113\n"
     ]
    }
   ],
   "source": [
    "pred = Optimized_Model.predict(X_validation)\n",
    "\n",
    "print('the accuracy is : ', metrics.accuracy_score(Y_validation, pred, normalize=True, sample_weight=None))\n",
    "print('the roc-auc score is :', metrics.roc_auc_score(Y_validation, pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Train the model on the entire dataset :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Pipeline(steps=[('Scaler', StandardScaler(copy=True, with_mean=True, with_std=True)), ('XGB', XGBClassifier(base_score=0.5, colsample_bylevel=1, colsample_bytree=1,\n",
       "       gamma=0.0, learning_rate=0.1, max_delta_step=0, max_depth=10,\n",
       "       min_child_weight=2, missing=None, n_estimators=100, nthread=-1,\n",
       "       objective='binary:logistic', reg_alpha=0, reg_lambda=1,\n",
       "       scale_pos_weight=1, seed=0, silent=True, subsample=1))])"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Optimized_Model.fit(Train_Input.iloc[:,1:], Train_Outputs.iloc[:,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# save the model to disk\n",
    "filename = 'finalized_model_PredictionOfCrudeOilTrend.sav'\n",
    "dump(Optimized_Model, open(filename, 'wb'))"
   ]
  }
 ],
 "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": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}