Feature Selection in ML using Python. Part 2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Univariate"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using Mutual Information Gain"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Classification Problem"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import accuracy_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.feature_selection import VarianceThreshold, mutual_info_classif, mutual_info_regression\n",
    "from sklearn.feature_selection import SelectKBest, SelectPercentile"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ID</th>\n",
       "      <th>var3</th>\n",
       "      <th>var15</th>\n",
       "      <th>imp_ent_var16_ult1</th>\n",
       "      <th>imp_op_var39_comer_ult1</th>\n",
       "      <th>imp_op_var39_comer_ult3</th>\n",
       "      <th>imp_op_var40_comer_ult1</th>\n",
       "      <th>imp_op_var40_comer_ult3</th>\n",
       "      <th>imp_op_var40_efect_ult1</th>\n",
       "      <th>imp_op_var40_efect_ult3</th>\n",
       "      <th>...</th>\n",
       "      <th>saldo_medio_var33_hace2</th>\n",
       "      <th>saldo_medio_var33_hace3</th>\n",
       "      <th>saldo_medio_var33_ult1</th>\n",
       "      <th>saldo_medio_var33_ult3</th>\n",
       "      <th>saldo_medio_var44_hace2</th>\n",
       "      <th>saldo_medio_var44_hace3</th>\n",
       "      <th>saldo_medio_var44_ult1</th>\n",
       "      <th>saldo_medio_var44_ult3</th>\n",
       "      <th>var38</th>\n",
       "      <th>TARGET</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>23</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>39205.170000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>49278.030000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4</td>\n",
       "      <td>2</td>\n",
       "      <td>23</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>67333.770000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>8</td>\n",
       "      <td>2</td>\n",
       "      <td>37</td>\n",
       "      <td>0.0</td>\n",
       "      <td>195.0</td>\n",
       "      <td>195.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>64007.970000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>10</td>\n",
       "      <td>2</td>\n",
       "      <td>39</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>117310.979016</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 371 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   ID  var3  var15  imp_ent_var16_ult1  imp_op_var39_comer_ult1  \\\n",
       "0   1     2     23                 0.0                      0.0   \n",
       "1   3     2     34                 0.0                      0.0   \n",
       "2   4     2     23                 0.0                      0.0   \n",
       "3   8     2     37                 0.0                    195.0   \n",
       "4  10     2     39                 0.0                      0.0   \n",
       "\n",
       "   imp_op_var39_comer_ult3  imp_op_var40_comer_ult1  imp_op_var40_comer_ult3  \\\n",
       "0                      0.0                      0.0                      0.0   \n",
       "1                      0.0                      0.0                      0.0   \n",
       "2                      0.0                      0.0                      0.0   \n",
       "3                    195.0                      0.0                      0.0   \n",
       "4                      0.0                      0.0                      0.0   \n",
       "\n",
       "   imp_op_var40_efect_ult1  imp_op_var40_efect_ult3  ...  \\\n",
       "0                      0.0                      0.0  ...   \n",
       "1                      0.0                      0.0  ...   \n",
       "2                      0.0                      0.0  ...   \n",
       "3                      0.0                      0.0  ...   \n",
       "4                      0.0                      0.0  ...   \n",
       "\n",
       "   saldo_medio_var33_hace2  saldo_medio_var33_hace3  saldo_medio_var33_ult1  \\\n",
       "0                      0.0                      0.0                     0.0   \n",
       "1                      0.0                      0.0                     0.0   \n",
       "2                      0.0                      0.0                     0.0   \n",
       "3                      0.0                      0.0                     0.0   \n",
       "4                      0.0                      0.0                     0.0   \n",
       "\n",
       "   saldo_medio_var33_ult3  saldo_medio_var44_hace2  saldo_medio_var44_hace3  \\\n",
       "0                     0.0                      0.0                      0.0   \n",
       "1                     0.0                      0.0                      0.0   \n",
       "2                     0.0                      0.0                      0.0   \n",
       "3                     0.0                      0.0                      0.0   \n",
       "4                     0.0                      0.0                      0.0   \n",
       "\n",
       "   saldo_medio_var44_ult1  saldo_medio_var44_ult3          var38  TARGET  \n",
       "0                     0.0                     0.0   39205.170000       0  \n",
       "1                     0.0                     0.0   49278.030000       0  \n",
       "2                     0.0                     0.0   67333.770000       0  \n",
       "3                     0.0                     0.0   64007.970000       0  \n",
       "4                     0.0                     0.0  117310.979016       0  \n",
       "\n",
       "[5 rows x 371 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv('santander-train.csv')\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = data.drop(labels=['TARGET'], axis=1)\n",
    "y = data['TARGET']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(76020, 370)\n",
      "(76020,)\n"
     ]
    }
   ],
   "source": [
    "print(X.shape)\n",
    "print(y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "### Removing Constants, Quasi Constants and Duplicate Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "VarianceThreshold(threshold=0.01)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "constant_filter = VarianceThreshold(threshold=0.01)\n",
    "constant_filter.fit(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_filter = constant_filter.transform(X_train)\n",
    "X_test_filter = constant_filter.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(57015, 272)\n",
      "(19005, 272)\n"
     ]
    }
   ],
   "source": [
    "print(X_train_filter.shape)\n",
    "print(X_test_filter.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_T = X_train_filter.T\n",
    "X_test_T = X_test_filter.T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'numpy.ndarray'>\n",
      "<class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "print(type(X_train_T))\n",
    "print(type(X_test_T))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_T = pd.DataFrame(X_train_T)\n",
    "X_test_T = pd.DataFrame(X_test_T)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "17"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_T.duplicated().sum()  ## Number of duplicate Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "duplicated_features = X_train_T.duplicated()\n",
    "features_to_keep = [not index for index in duplicated_features]\n",
    "X_train_unique = X_train_T[features_to_keep].T\n",
    "X_test_unique = X_test_T[features_to_keep].T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(57015, 255)\n",
      "(19005, 255)\n"
     ]
    }
   ],
   "source": [
    "print(X_train_unique.shape)\n",
    "print(X_test_unique.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "### Calculating Mutual Information gain between each feature and Target "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "mi = mutual_info_classif(X_train_unique, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "print(type(mi))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([8.94550619e-04, 9.03515337e-04, 1.33003603e-02, 0.00000000e+00,\n",
       "       6.71817795e-05, 3.62315820e-04, 3.56386717e-04, 1.60291120e-04,\n",
       "       0.00000000e+00, 2.46436083e-04, 6.86225795e-05, 1.12317065e-03,\n",
       "       3.62206502e-04, 1.09594126e-03, 1.07944044e-03, 6.76562325e-04,\n",
       "       0.00000000e+00, 1.21252236e-03, 8.01205894e-04, 1.28622830e-03,\n",
       "       3.09868030e-04, 1.07841185e-02, 1.44185851e-02, 1.79879081e-03,\n",
       "       0.00000000e+00, 2.23672560e-03, 3.35486473e-04, 2.72095319e-04,\n",
       "       1.61259831e-03, 1.08256862e-03, 1.64785788e-03, 5.74525794e-04,\n",
       "       4.17045670e-04, 6.37153784e-04, 8.97948725e-04, 0.00000000e+00,\n",
       "       3.05310064e-04, 1.36889371e-04, 5.79055563e-04, 1.71924198e-02,\n",
       "       4.02114844e-04, 3.33982802e-04, 1.09111320e-02, 1.45835008e-03,\n",
       "       1.10809615e-02, 0.00000000e+00, 3.36505872e-04, 1.38071446e-02,\n",
       "       1.28546019e-02, 1.71389716e-02, 1.65078615e-03, 2.56898609e-04,\n",
       "       2.66344410e-03, 2.86981737e-04, 0.00000000e+00, 4.69442039e-04,\n",
       "       1.13041893e-03, 0.00000000e+00, 7.86640271e-04, 1.70556503e-03,\n",
       "       0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 7.02852020e-04,\n",
       "       2.59337934e-04, 2.17984241e-04, 3.55969372e-04, 1.36381947e-03,\n",
       "       4.33406465e-04, 1.12327981e-04, 4.24967970e-04, 0.00000000e+00,\n",
       "       1.51190591e-04, 0.00000000e+00, 1.74762631e-04, 3.36673869e-04,\n",
       "       5.28302955e-04, 0.00000000e+00, 4.11256867e-04, 0.00000000e+00,\n",
       "       8.37150761e-04, 6.12181389e-05, 8.81004180e-03, 1.70661361e-02,\n",
       "       1.58141286e-03, 1.63273586e-04, 1.10120933e-03, 1.51347192e-02,\n",
       "       0.00000000e+00, 0.00000000e+00, 9.24279858e-03, 0.00000000e+00,\n",
       "       7.21926360e-04, 9.17146764e-03, 1.08931605e-02, 1.58947452e-02,\n",
       "       2.99796889e-04, 0.00000000e+00, 7.10621988e-04, 9.88607499e-03,\n",
       "       0.00000000e+00, 6.94402983e-04, 2.70312362e-03, 2.85976709e-04,\n",
       "       9.12301025e-04, 5.57854315e-04, 0.00000000e+00, 0.00000000e+00,\n",
       "       1.36899338e-03, 4.78789824e-06, 6.13694798e-04, 2.49576880e-04,\n",
       "       0.00000000e+00, 8.32853521e-04, 1.30671933e-02, 6.13095374e-04,\n",
       "       0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n",
       "       0.00000000e+00, 1.25318879e-02, 6.61967572e-05, 1.02598536e-02,\n",
       "       5.04938984e-04, 0.00000000e+00, 1.16563919e-04, 0.00000000e+00,\n",
       "       0.00000000e+00, 2.62369217e-06, 2.48296850e-04, 0.00000000e+00,\n",
       "       8.46639206e-04, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n",
       "       0.00000000e+00, 6.24555574e-04, 6.85222109e-04, 1.00846981e-03,\n",
       "       1.09509351e-03, 0.00000000e+00, 1.22193233e-03, 0.00000000e+00,\n",
       "       0.00000000e+00, 1.38676361e-03, 1.10415313e-04, 3.17305308e-05,\n",
       "       0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 3.57855573e-04,\n",
       "       2.78015303e-04, 0.00000000e+00, 8.34802673e-04, 0.00000000e+00,\n",
       "       2.74338514e-04, 1.48426132e-03, 0.00000000e+00, 3.94470684e-04,\n",
       "       6.30666368e-04, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n",
       "       0.00000000e+00, 1.98242308e-04, 0.00000000e+00, 0.00000000e+00,\n",
       "       7.24982937e-04, 4.44622599e-04, 8.82297306e-04, 0.00000000e+00,\n",
       "       0.00000000e+00, 7.81512815e-04, 8.08217421e-04, 0.00000000e+00,\n",
       "       2.07510665e-04, 5.78688541e-04, 0.00000000e+00, 1.25891611e-03,\n",
       "       2.33440087e-04, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n",
       "       1.02841722e-03, 4.44587906e-04, 6.91077652e-04, 2.56274343e-04,\n",
       "       9.05277993e-04, 1.53275968e-02, 0.00000000e+00, 1.19634945e-03,\n",
       "       1.21928916e-03, 1.14931690e-03, 9.04509167e-03, 0.00000000e+00,\n",
       "       5.39012109e-04, 3.67519526e-04, 0.00000000e+00, 0.00000000e+00,\n",
       "       0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n",
       "       5.79359027e-04, 0.00000000e+00, 5.83592068e-04, 8.57035515e-04,\n",
       "       7.31193179e-04, 3.44043233e-04, 0.00000000e+00, 7.41776252e-04,\n",
       "       0.00000000e+00, 1.43144573e-03, 4.61988959e-04, 2.13917456e-03,\n",
       "       1.56955098e-04, 1.13475580e-02, 8.79886230e-03, 1.10790139e-02,\n",
       "       9.10175746e-03, 6.13253975e-04, 0.00000000e+00, 0.00000000e+00,\n",
       "       4.11728641e-04, 4.65797370e-04, 7.11212036e-04, 0.00000000e+00,\n",
       "       2.56417526e-03, 1.96889660e-03, 9.46400418e-04, 1.22423913e-03,\n",
       "       1.03713365e-03, 2.92990252e-05, 0.00000000e+00, 1.20016939e-03,\n",
       "       0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n",
       "       0.00000000e+00, 6.98890623e-05, 1.13804848e-03, 7.36480017e-04,\n",
       "       5.27914292e-04, 5.38690590e-04, 0.00000000e+00, 0.00000000e+00,\n",
       "       9.29945831e-04, 0.00000000e+00, 0.00000000e+00, 5.60398244e-04,\n",
       "       0.00000000e+00, 7.00957261e-04, 1.74141823e-03])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "255\n"
     ]
    }
   ],
   "source": [
    "print(len(mi))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0      0.000895\n",
       "1      0.000904\n",
       "2      0.013300\n",
       "3      0.000000\n",
       "4      0.000067\n",
       "5      0.000362\n",
       "6      0.000356\n",
       "7      0.000160\n",
       "8      0.000000\n",
       "9      0.000246\n",
       "10     0.000069\n",
       "11     0.001123\n",
       "12     0.000362\n",
       "13     0.001096\n",
       "14     0.001079\n",
       "15     0.000677\n",
       "16     0.000000\n",
       "17     0.001213\n",
       "18     0.000801\n",
       "19     0.001286\n",
       "20     0.000310\n",
       "21     0.010784\n",
       "22     0.014419\n",
       "23     0.001799\n",
       "24     0.000000\n",
       "25     0.002237\n",
       "26     0.000335\n",
       "27     0.000272\n",
       "28     0.001613\n",
       "29     0.001083\n",
       "         ...   \n",
       "241    0.000466\n",
       "242    0.000711\n",
       "243    0.000000\n",
       "244    0.002564\n",
       "245    0.001969\n",
       "246    0.000946\n",
       "247    0.001224\n",
       "248    0.001037\n",
       "249    0.000029\n",
       "250    0.000000\n",
       "251    0.001200\n",
       "252    0.000000\n",
       "253    0.000000\n",
       "255    0.000000\n",
       "256    0.000000\n",
       "257    0.000000\n",
       "258    0.000070\n",
       "259    0.001138\n",
       "260    0.000736\n",
       "261    0.000528\n",
       "262    0.000539\n",
       "263    0.000000\n",
       "264    0.000000\n",
       "265    0.000930\n",
       "266    0.000000\n",
       "267    0.000000\n",
       "268    0.000560\n",
       "269    0.000000\n",
       "270    0.000701\n",
       "271    0.001741\n",
       "Length: 255, dtype: float64"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mi = pd.Series(mi, index=X_train_unique.columns)\n",
    "mi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "41     0.017192\n",
       "52     0.017139\n",
       "88     0.017066\n",
       "103    0.015895\n",
       "205    0.015328\n",
       "93     0.015135\n",
       "22     0.014419\n",
       "50     0.013807\n",
       "2      0.013300\n",
       "123    0.013067\n",
       "51     0.012855\n",
       "130    0.012532\n",
       "233    0.011348\n",
       "47     0.011081\n",
       "235    0.011079\n",
       "45     0.010911\n",
       "102    0.010893\n",
       "21     0.010784\n",
       "132    0.010260\n",
       "107    0.009886\n",
       "97     0.009243\n",
       "100    0.009171\n",
       "236    0.009102\n",
       "210    0.009045\n",
       "87     0.008810\n",
       "234    0.008799\n",
       "110    0.002703\n",
       "55     0.002663\n",
       "244    0.002564\n",
       "25     0.002237\n",
       "         ...   \n",
       "134    0.000000\n",
       "64     0.000000\n",
       "105    0.000000\n",
       "65     0.000000\n",
       "228    0.000000\n",
       "76     0.000000\n",
       "78     0.000000\n",
       "82     0.000000\n",
       "84     0.000000\n",
       "226    0.000000\n",
       "94     0.000000\n",
       "95     0.000000\n",
       "197    0.000000\n",
       "221    0.000000\n",
       "219    0.000000\n",
       "218    0.000000\n",
       "129    0.000000\n",
       "108    0.000000\n",
       "217    0.000000\n",
       "216    0.000000\n",
       "114    0.000000\n",
       "115    0.000000\n",
       "215    0.000000\n",
       "120    0.000000\n",
       "214    0.000000\n",
       "125    0.000000\n",
       "126    0.000000\n",
       "127    0.000000\n",
       "128    0.000000\n",
       "136    0.000000\n",
       "Length: 255, dtype: float64"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mi.sort_values(ascending=False, inplace=True)\n",
    "mi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x20b63c310f0>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "mi.plot.bar(figsize = (12, 8))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "### From the graph we can see that at rhe end values gave MI==0 which means they are not putting much information indivisually but in a group they may put information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "### Selecting top 10 percentile of the features "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Int64Index([  2,  21,  22,  41,  45,  47,  50,  51,  52,  87,  88,  93,  97,\n",
       "            100, 102, 103, 107, 123, 130, 132, 205, 210, 233, 234, 235, 236],\n",
       "           dtype='int64')"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sel = SelectPercentile(score_func=mutual_info_classif, percentile=10)\n",
    "sel.fit(X_train_unique, y_train)\n",
    "X_train_unique.columns[sel.get_support()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_mi = sel.transform(X_train_unique)\n",
    "X_test_mi = sel.transform(X_test_unique)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(57015, 26)\n",
      "(19005, 26)\n"
     ]
    }
   ],
   "source": [
    "print(X_train_mi.shape)\n",
    "print(X_test_mi.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.959431728492502\n",
      "Wall time: 6.5 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "model = RandomForestClassifier(n_estimators=100, max_depth=4)\n",
    "model.fit(X_train, y_train)\n",
    "y_pred = model.predict(X_test)\n",
    "print(accuracy_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.959431728492502\n",
      "Wall time: 2.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "model = RandomForestClassifier(n_estimators=100, max_depth=4)\n",
    "model.fit(X_train_mi, y_train)\n",
    "y_pred = model.predict(X_test_mi)\n",
    "print(accuracy_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Regression Problem"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_boston\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.metrics import r2_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "boston = load_boston()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      ".. _boston_dataset:\n",
      "\n",
      "Boston house prices dataset\n",
      "---------------------------\n",
      "\n",
      "**Data Set Characteristics:**  \n",
      "\n",
      "    :Number of Instances: 506 \n",
      "\n",
      "    :Number of Attributes: 13 numeric/categorical predictive. Median Value (attribute 14) is usually the target.\n",
      "\n",
      "    :Attribute Information (in order):\n",
      "        - CRIM     per capita crime rate by town\n",
      "        - ZN       proportion of residential land zoned for lots over 25,000 sq.ft.\n",
      "        - INDUS    proportion of non-retail business acres per town\n",
      "        - CHAS     Charles River dummy variable (= 1 if tract bounds river; 0 otherwise)\n",
      "        - NOX      nitric oxides concentration (parts per 10 million)\n",
      "        - RM       average number of rooms per dwelling\n",
      "        - AGE      proportion of owner-occupied units built prior to 1940\n",
      "        - DIS      weighted distances to five Boston employment centres\n",
      "        - RAD      index of accessibility to radial highways\n",
      "        - TAX      full-value property-tax rate per $10,000\n",
      "        - PTRATIO  pupil-teacher ratio by town\n",
      "        - B        1000(Bk - 0.63)^2 where Bk is the proportion of blacks by town\n",
      "        - LSTAT    % lower status of the population\n",
      "        - MEDV     Median value of owner-occupied homes in $1000's\n",
      "\n",
      "    :Missing Attribute Values: None\n",
      "\n",
      "    :Creator: Harrison, D. and Rubinfeld, D.L.\n",
      "\n",
      "This is a copy of UCI ML housing dataset.\n",
      "https://archive.ics.uci.edu/ml/machine-learning-databases/housing/\n",
      "\n",
      "\n",
      "This dataset was taken from the StatLib library which is maintained at Carnegie Mellon University.\n",
      "\n",
      "The Boston house-price data of Harrison, D. and Rubinfeld, D.L. 'Hedonic\n",
      "prices and the demand for clean air', J. Environ. Economics & Management,\n",
      "vol.5, 81-102, 1978.   Used in Belsley, Kuh & Welsch, 'Regression diagnostics\n",
      "...', Wiley, 1980.   N.B. Various transformations are used in the table on\n",
      "pages 244-261 of the latter.\n",
      "\n",
      "The Boston house-price data has been used in many machine learning papers that address regression\n",
      "problems.   \n",
      "     \n",
      ".. topic:: References\n",
      "\n",
      "   - Belsley, Kuh & Welsch, 'Regression diagnostics: Identifying Influential Data and Sources of Collinearity', Wiley, 1980. 244-261.\n",
      "   - Quinlan,R. (1993). Combining Instance-Based and Model-Based Learning. In Proceedings on the Tenth International Conference of Machine Learning, 236-243, University of Massachusetts, Amherst. Morgan Kaufmann.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(boston.DESCR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(506, 13)\n"
     ]
    }
   ],
   "source": [
    "X = pd.DataFrame(data = boston.data, columns= boston.feature_names)\n",
    "print(X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CRIM</th>\n",
       "      <th>ZN</th>\n",
       "      <th>INDUS</th>\n",
       "      <th>CHAS</th>\n",
       "      <th>NOX</th>\n",
       "      <th>RM</th>\n",
       "      <th>AGE</th>\n",
       "      <th>DIS</th>\n",
       "      <th>RAD</th>\n",
       "      <th>TAX</th>\n",
       "      <th>PTRATIO</th>\n",
       "      <th>B</th>\n",
       "      <th>LSTAT</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.00632</td>\n",
       "      <td>18.0</td>\n",
       "      <td>2.31</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.538</td>\n",
       "      <td>6.575</td>\n",
       "      <td>65.2</td>\n",
       "      <td>4.0900</td>\n",
       "      <td>1.0</td>\n",
       "      <td>296.0</td>\n",
       "      <td>15.3</td>\n",
       "      <td>396.90</td>\n",
       "      <td>4.98</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.02731</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7.07</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.469</td>\n",
       "      <td>6.421</td>\n",
       "      <td>78.9</td>\n",
       "      <td>4.9671</td>\n",
       "      <td>2.0</td>\n",
       "      <td>242.0</td>\n",
       "      <td>17.8</td>\n",
       "      <td>396.90</td>\n",
       "      <td>9.14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.02729</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7.07</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.469</td>\n",
       "      <td>7.185</td>\n",
       "      <td>61.1</td>\n",
       "      <td>4.9671</td>\n",
       "      <td>2.0</td>\n",
       "      <td>242.0</td>\n",
       "      <td>17.8</td>\n",
       "      <td>392.83</td>\n",
       "      <td>4.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.03237</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.18</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.458</td>\n",
       "      <td>6.998</td>\n",
       "      <td>45.8</td>\n",
       "      <td>6.0622</td>\n",
       "      <td>3.0</td>\n",
       "      <td>222.0</td>\n",
       "      <td>18.7</td>\n",
       "      <td>394.63</td>\n",
       "      <td>2.94</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.06905</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.18</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.458</td>\n",
       "      <td>7.147</td>\n",
       "      <td>54.2</td>\n",
       "      <td>6.0622</td>\n",
       "      <td>3.0</td>\n",
       "      <td>222.0</td>\n",
       "      <td>18.7</td>\n",
       "      <td>396.90</td>\n",
       "      <td>5.33</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      CRIM    ZN  INDUS  CHAS    NOX     RM   AGE     DIS  RAD    TAX  \\\n",
       "0  0.00632  18.0   2.31   0.0  0.538  6.575  65.2  4.0900  1.0  296.0   \n",
       "1  0.02731   0.0   7.07   0.0  0.469  6.421  78.9  4.9671  2.0  242.0   \n",
       "2  0.02729   0.0   7.07   0.0  0.469  7.185  61.1  4.9671  2.0  242.0   \n",
       "3  0.03237   0.0   2.18   0.0  0.458  6.998  45.8  6.0622  3.0  222.0   \n",
       "4  0.06905   0.0   2.18   0.0  0.458  7.147  54.2  6.0622  3.0  222.0   \n",
       "\n",
       "   PTRATIO       B  LSTAT  \n",
       "0     15.3  396.90   4.98  \n",
       "1     17.8  396.90   9.14  \n",
       "2     17.8  392.83   4.03  \n",
       "3     18.7  394.63   2.94  \n",
       "4     18.7  396.90   5.33  "
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[24.  21.6 34.7 33.4 36.2]\n"
     ]
    }
   ],
   "source": [
    "y = boston.target\n",
    "print(y[: 5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "mi = mutual_info_regression(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "13\n"
     ]
    }
   ],
   "source": [
    "print(len(mi))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2.99951431e-01, 1.43247778e-01, 4.38566164e-01, 3.85464658e-04,\n",
       "       4.13610240e-01, 5.31037042e-01, 3.08638119e-01, 2.64890894e-01,\n",
       "       1.84901879e-01, 3.41865725e-01, 4.33617543e-01, 1.86900020e-01,\n",
       "       6.43439919e-01])"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "mi = pd.Series(mi, index=X.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "CRIM       0.299951\n",
       "ZN         0.143248\n",
       "INDUS      0.438566\n",
       "CHAS       0.000385\n",
       "NOX        0.413610\n",
       "RM         0.531037\n",
       "AGE        0.308638\n",
       "DIS        0.264891\n",
       "RAD        0.184902\n",
       "TAX        0.341866\n",
       "PTRATIO    0.433618\n",
       "B          0.186900\n",
       "LSTAT      0.643440\n",
       "dtype: float64"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "mi.sort_values(ascending=False, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LSTAT      0.643440\n",
       "RM         0.531037\n",
       "INDUS      0.438566\n",
       "PTRATIO    0.433618\n",
       "NOX        0.413610\n",
       "TAX        0.341866\n",
       "AGE        0.308638\n",
       "CRIM       0.299951\n",
       "DIS        0.264891\n",
       "B          0.186900\n",
       "RAD        0.184902\n",
       "ZN         0.143248\n",
       "CHAS       0.000385\n",
       "dtype: float64"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x20b699e8e80>"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEcCAYAAAA88/RnAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAGtVJREFUeJzt3XuUXVWB5/Hvj0SURrRVyraFYALEdhBRNKKz1PbdwqCAikrabsVRgz3EF2obHQeXqN2Mb20ZNdo0jksEHwMESUv7QEftRlOKCgmi6YgQacdoo2L7wMBv/tin4OZyU3Wrsu+tyub3WatW3fOos/etuvW75+6z9z6yTUREtGWP+a5ARETUl3CPiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokEJ94iIBiXcIyIatHi+Ct533329dOnS+So+ImK39I1vfOOntidm2m/ewn3p0qVMTk7OV/EREbslST8cZr80y0RENCjhHhHRoIR7RESDEu4REQ1KuEdENCjhHhHRoIR7RESDEu4REQ2at0FM01m65qJZ7X/16UePqCYREbunnLlHRDQo4R4R0aCEe0REgxLuERENSrhHRDQo4R4R0aCEe0REgxLuERENSrhHRDQo4R4R0aChwl3SkZKukrRZ0pqd7PNMSZskbZR0dt1qRkTEbMw4t4ykRcAZwBOBrcAGSetsb+rZZznwGuARtq+XdM9RVTgiImY2zJn7EcBm21ts3wicAxzbt88LgTNsXw9g+yd1qxkREbMxTLjvB1zbs7y1W9frvsB9JX1V0qWSjhx0IEmrJE1Kmty2bdvcahwRETMaJtw1YJ37lhcDy4HHACuBD0n6w9v8kL3W9grbKyYmJmZb14iIGNIw4b4VWNKzvD9w3YB9LrD9e9s/AK6ihH1ERMyDYcJ9A7Bc0jJJewInAOv69jkfeCyApH0pzTRbalY0IiKGN2O4294OrAYuBq4EPm57o6TTJB3T7XYx8DNJm4BLgFfZ/tmoKh0REdMb6jZ7ttcD6/vWndrz2MAp3VdERMyzjFCNiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokEJ94iIBiXcIyIalHCPiGjQUHdiatHSNRfN+meuPv3oEdQkIqK+nLlHRDQo4R4R0aCEe0REgxLuERENSrhHRDRoqHCXdKSkqyRtlrRmwPYTJW2T9K3u6wX1qxoREcOasSukpEXAGcATga3ABknrbG/q2/Vc26tHUMeIiJilYc7cjwA2295i+0bgHODY0VYrIiJ2xTDhvh9wbc/y1m5dv6dL+o6kT0paUqV2ERExJ8OEuwasc9/yhcBS24cBnwM+PPBA0ipJk5Imt23bNruaRkTE0IYJ961A75n4/sB1vTvY/pnt33WLHwQeMuhAttfaXmF7xcTExFzqGxERQxgm3DcAyyUtk7QncAKwrncHSX/cs3gMcGW9KkZExGzN2FvG9nZJq4GLgUXAmbY3SjoNmLS9DniJpGOA7cC/AyeOsM4RETGDoWaFtL0eWN+37tSex68BXlO3ahERMVcZoRoR0aCEe0REgxLuERENSrhHRDQo4R4R0aCEe0REgxLuERENSrhHRDQo4R4R0aCEe0REgxLuERENSrhHRDQo4R4R0aCEe0REg4aa8jfmbumai2b9M1effvQIahIRtyc5c4+IaFDO3BuRTwgR0Stn7hERDUq4R0Q0KOEeEdGghHtERIMS7hERDUpvmZiV2fbKSY+ciPkx1Jm7pCMlXSVps6Q10+x3vCRLWlGvihERMVszhrukRcAZwFHAIcBKSYcM2G8f4CXA12pXMiIiZmeYM/cjgM22t9i+ETgHOHbAfm8E3gL8tmL9IiJiDoYJ9/2Aa3uWt3brbiHpcGCJ7U9XrFtERMzRMOGuAet8y0ZpD+CdwCtmPJC0StKkpMlt27YNX8uIiJiVYcJ9K7CkZ3l/4Lqe5X2AQ4EvSroaeDiwbtBFVdtrba+wvWJiYmLutY6IiGkNE+4bgOWSlknaEzgBWDe10fYvbO9re6ntpcClwDG2J0dS44iImNGM4W57O7AauBi4Evi47Y2STpN0zKgrGBERszfUICbb64H1fetO3cm+j9n1akVExK7I9AMREQ1KuEdENCjhHhHRoIR7RESDEu4REQ1KuEdENCjhHhHRoIR7RESDEu4REQ1KuEdENCjhHhHRoIR7RESDEu4REQ0aalbIiHFauuaiWf/M1acfPYKaROy+cuYeEdGghHtERIMS7hERDUq4R0Q0KOEeEdGghHtERIMS7hERDUq4R0Q0KOEeEdGghHtERIOGmn5A0pHAu4FFwIdsn963/UXAycBNwK+AVbY3Va5rRFWZ5iBaNuOZu6RFwBnAUcAhwEpJh/TtdrbtB9h+EPAW4B3VaxoREUMbplnmCGCz7S22bwTOAY7t3cH2L3sW9wZcr4oRETFbwzTL7Adc27O8FXhY/06STgZOAfYEHjfoQJJWAasADjjggNnWNWK3lOafmA/DnLlrwLrbnJnbPsP2QcCrgdcNOpDttbZX2F4xMTExu5pGRMTQhgn3rcCSnuX9geum2f8c4LhdqVREROyaYcJ9A7Bc0jJJewInAOt6d5C0vGfxaOD79aoYERGzNWObu+3tklYDF1O6Qp5pe6Ok04BJ2+uA1ZKeAPweuB547igrHRER0xuqn7vt9cD6vnWn9jx+aeV6RUTELsgI1YiIBiXcIyIalHCPiGhQwj0iokEJ94iIBg3VWyYiFr7ZTnOQKQ7aljP3iIgGJdwjIhqUcI+IaFDCPSKiQQn3iIgGJdwjIhqUcI+IaFDCPSKiQQn3iIgGJdwjIhqUcI+IaFDCPSKiQQn3iIgGJdwjIhqUcI+IaFDCPSKiQQn3iIgGDRXuko6UdJWkzZLWDNh+iqRNkr4j6fOS7lO/qhERMawZw13SIuAM4CjgEGClpEP6drsMWGH7MOCTwFtqVzQiIoY3zJn7EcBm21ts3wicAxzbu4PtS2z/ulu8FNi/bjUjImI2hgn3/YBre5a3dut25vnAPw7aIGmVpElJk9u2bRu+lhERMSvDhLsGrPPAHaW/AFYAbx203fZa2ytsr5iYmBi+lhERMSuLh9hnK7CkZ3l/4Lr+nSQ9AfjvwKNt/65O9SIiYi6GOXPfACyXtEzSnsAJwLreHSQdDnwAOMb2T+pXMyIiZmPGM3fb2yWtBi4GFgFn2t4o6TRg0vY6SjPMnYFPSAK4xvYxI6x3RMyDpWsumvXPXH360SOoScxkmGYZbK8H1vetO7Xn8RMq1ysibsfyJrLrMkI1IqJBCfeIiAYl3CMiGpRwj4hoUMI9IqJBCfeIiAYl3CMiGpRwj4hoUMI9IqJBCfeIiAYl3CMiGpRwj4ho0FATh0VEtKjlCcpy5h4R0aCEe0REgxLuERENSrhHRDQo4R4R0aCEe0REgxLuERENSrhHRDQo4R4R0aCEe0REg4YKd0lHSrpK0mZJawZs/1NJ35S0XdLx9asZERGzMWO4S1oEnAEcBRwCrJR0SN9u1wAnAmfXrmBERMzeMBOHHQFstr0FQNI5wLHApqkdbF/dbbt5BHWMiIhZGqZZZj/g2p7lrd26WZO0StKkpMlt27bN5RARETGEYcJdA9Z5LoXZXmt7he0VExMTczlEREQMYZhw3wos6VneH7huNNWJiIgahgn3DcByScsk7QmcAKwbbbUiImJXzBjutrcDq4GLgSuBj9veKOk0SccASHqopK3AM4APSNo4ykpHRMT0hrrNnu31wPq+daf2PN5Aaa6JiIgFICNUIyIalHCPiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokEJ94iIBiXcIyIalHCPiGhQwj0iokFDhbukIyVdJWmzpDUDtt9R0rnd9q9JWlq7ohERMbwZw13SIuAM4CjgEGClpEP6dns+cL3tg4F3Av+zdkUjImJ4w5y5HwFstr3F9o3AOcCxffscC3y4e/xJ4PGSVK+aERExG7I9/Q7S8cCRtl/QLf8l8DDbq3v2uaLbZ2u3/K/dPj/tO9YqYFW3+CfAVbOs777AT2fca9eNo5yWnktr5bT0XForp6XnMtdy7mN7YqadFg9xoEFn4P3vCMPsg+21wNohyhxcEWnS9oq5/vxCKqel59JaOS09l9bKaem5jLqcYZpltgJLepb3B67b2T6SFgN3Bf69RgUjImL2hgn3DcByScsk7QmcAKzr22cd8Nzu8fHAFzxTe09ERIzMjM0ytrdLWg1cDCwCzrS9UdJpwKTtdcDfAx+RtJlyxn7CiOo75yadBVhOS8+ltXJaei6tldPScxlpOTNeUI2IiN1PRqhGRDQo4R4R0aCEe0REgxLusSBJuvM02w4aZ10idkcLNtwl/c1812FUJN1B0uGS7jnfdZkLSU/fyfo9Jf2PSsV8W9Iz+45/J0lvAj5TqQwk3a/n8R37tj28YjlPm+6rVjkDyr2HpKdKesioyugpa99MO7JwLNjeMpK+afvBYyjnPdNtt/2SCmW8H/i7rgvpXYF/AW4C7g680vbHdrWMnrIOBf6aMsmbgU3A221/p2IZFwM3A//N9g+6dUdRJo37jO2XVSjjIOC9lO66fwXcH3gbcD7wBtu/2tUyunJueZ31v+ZqvgYl3Qx8q/uCHUd12/Z/rVTOp4E1tq+Q9MfAN4FJ4CBgre13VSrn4cDplK7PbwQ+QhlKvwfwHNtV3oAlPRZ4MWW6EoArgffa/mKN43dlPGe67bb/d4UyXgh80fb3uzfAM4GnA1cDJ9r+5q6W0W+Y6QfmyyJJd2Pw1AbYrjUC9kXAFcDHKSNvR3Hm8SjbL+oePw/4nu3jJN0L+EegSrhLOpYSgH/bfRfwEOBTkl5p+4Ia5dh+kqSVwOcknQ0cCkwAz7L97Upl/CtwlKRXAd8Ffgw8yfbGGsfvoZ08HrS8K54OPAs4DLgA+JjtzRWPP2WZ7Su6x88DPmv7OZL2Ab4KVAl3yhvvaymj0b8AHGX70u6T0Meo8OlK0tFdOacBb6D8PR4MnClpte31u1pG56GDigeeAuwH7HK4Ay8Fzuoer6S8DpYBhwPvBh5VoYwd2V6QX8DvgC3ADwZ8balYzj0oAX8J8FngBcDdKj+Xy3oeX0R5p77NtgrlfBtYOmD9UuDblZ/TIuBNwK8o00/ct/LxFwOvATZTJps7H/g88CeVy/nmoMeDliuVtzfw55SA/wrw6MrH/1bP488DJwzaVrmcK/u2VXlNA18EHjhg/WHAl2r/bbpjC/gL4HLgXOCwEfy+zgZe2rNc/XVme0GfuW+yffioC7H9M+D9wPsl7Ud5V90o6dW2P1KpmJ9LejLwI+ARlPnvp+bh2atSGQB3sH11/0rbV0u6Q61CJD0S+F+UM8ElwKOBCyWdC7zZ9u8qFHMZ8CXgIbZ/AaztfocXSDrP9msqlAGwf9c0p57HdMv7VSqj12+BXwC/BA4A7lT5+NdKejHlDffBdGfQkvYCqr0GKM1yU37Tt61WW++9POCToO3vSPqjSmUAt/wvngi8AvgacLzt2c5aO52bu2ay64HHA2/u2VYzA26xkMN9rCQ9mBLsT6Q0lXyj4uFPAt4D3At4me0fd+sfTzmTr+X3kg6wfU3vSkn3AbZXLOddwAtsf71bPl/SZ4FTKZ8e7rfTnxzeibZ3+BvY/rSkzwGvq3D8Ka/qeTzZt61/ec66tuOVlPsjfA54t+1qx+/xfEozxhMozWQ/79Y/HPiHiuU8UNIvKW+Ce3WP6ZZrvWH9xxy3zYqkkynNJp+nTF3+w1rH7nEq5fW0CFjnrnlR0qMpLRTVLeQLqifaPmvA+jsBT7H9iUrlvAF4MuVCzTmUC4I1g3BsJB0HvAX4G8qbkyntiWuAV9s+v1I5e9i+eSfb/pPtK2uUs5PjPwL4c9snj6qMUeguqH6H0hRj+s5uXeHCfWsk/Rz4v4M2AY+0fbdK5dwM/ATYxo5/F1Eudh9WqZzFwD62r+9Z9wfAIts31Chjh/IWarj36m7192eUM58nAV+2fXylY99Meeec+mg59Qup9oeV9Hfs+KIxZYL+S2x/ZVeP31fWAykfLe9PeQ4bgbcN+ni7i+XcEzi5K2eqV84Ztn9Ss5yurAdR2qmfSbnm8inb76107EcCB7rrESHpk5ReTABvsv2FSuWcyDTNFbY/vLNtsyznwhnKOaZGOePQndXulO0vVSrnRZTmq0G/t2fZfkuNcvrKFPBYyuv6KbarNjPBAg93SX9KefJHA1+ntFcfaPvXFcu4z3Tba3xEk/TcAavvTgmrc12pe9q4dGfPZ1Ou/n+DW3sxPBd4tu2vVijjvpTZRVcCP6Nc3Hql7Wn/XnMo5/PAi21v6pYvp7S97g281vaRNcsbtXEF4nyStIRyofitlY53E+X6zl/a/lHftqpdsiU9jJJpT6VkwMmUZprrp/3BuZS1UMNd0lbgGuB9wPm2b5D0A9vLxlT+IsoL6KMjLGMv4J9rXTge11mbpEuBv7J9Wd/6BwEfsP2wCmXcDHwZeL67LoOSttg+cFeP3VfOBtsP7Vn+P7af1j3+qu1HVCpn7GfUkia6Y2+rfexxk7Qv8AzKm/1+wHm2X1np2JdROgicCpzS2+Qr6bIa/5+S3kw5mbuG0lX0PMqU6SPLs4V8QfVTwHGUvsE3SbqAelfhbyHpLpR3z/0oNx35LLAaeCVlwMnIwt32byoP6HtbzYNN4y79wQ5g+1tdf+oank45c79E0mco10NGMQbhD3sXpoK9U/Oj8rj+Nkh6PWXgj4A9JG2nDKI7bVx1qKF7LT2VcqZ7X0ogHmh7/8pF2fYHJX0J+Kik/wKc3LUQ1MqcVZR7Rr8P+LTt30oa7Zn1KPpX1vqivDgfB3yQ0o3wBsq7350rlnEBpXnhJMpAps9SPqI9aMTPbTFlkMmFFY951pj+LlcyYCwA5WPmd2v9frrvewPPBj4N/Jryz/FnFZ/LhcDRA9Y/GbhoTL/PR1Q81su71/CynnUHUm628/JxPJ+Kz+U33f/io7i1laHaGJeecnrHOiymjLy9CngYlfqgU3rJHEUZELWVMqL336Ze5yP5/c33H3AWv5w7UEaMnQ38tOJxL+/7A1xPuaJds+43UPo139Dz9f+6N5N7VyxnJIMhBpSzinL7xUcD+3Rfj6H0Dz5pVM+le/M4iXIbx1rP5eDuH/kfKGe7L+7e7L9HxYFZ3WtrJeUT4aHduicD/0zdgWyXAfsOWD9Rs5wxvc5e3r2mrqCMhj1oROF+m99L93reAtwwgvLuRLkd6ae6HDh7FL+/hdzmfpbtE3eybS/b/QMn5lrOyOYTGTdJ36UEyM6mbKg2f0U3oOivKb1loPTKeavtCysdv0pb55Bl3ZHy6WCq589G4PvASlfqcinpLMqAr69Tzgh/CPxnyjwwVbqoduVcYfvQ2W5byCQdSHldnwAsB15PaXP/XqXjHzfob9BNf3KS7dNrlLOTsvcBnuZKvaV2OPYCDvdxTRx2E7cOiBBltNivubUr5F0qlbOY8rFsaoDPJuBiV+xTL+kGyhn1oHC37cfVKmvUugvq79jZdts73bYLZR5OCZFRdLm8gjKU/eZurMZPgYN964C2Kqb7v9mdT1ymSHoAXbdY27vN1M+STplu+yhezwv5guofdP9sIz0Ltb2oxnGmI+nelLlr/o3ysVmUj+TvkPRY29dVKmrzOAJc0qnTbLbtN1YoZhFwZ0ZzEfUWO+lyKduPrVzUje4GfrlcTPte7WDvTI0c7Vdz5Oi8sX25yrTStSeQG7XejgYnAR/oWR7JGfZCPnNv6Sz0LMrEQe/qW/8Sytwpg/rBz6WcsTRlSHrFgNV7U4a+38P2Tm+0MYsyxvXJbVxdLn9NmQQNymv6oJ5lXGkUZEum6cn2CspEeMfOY/XmbFz/pwv5zH0sZ6Fj8vBB1w9sv0dSzcmJXj1oZe1BH7bf3nPsfSjzcjyP0l3x7Tv7uVka100fxtXl8oGUrpXX9q2/D2Wq6bitj1A6OPwLZbbWVwF7AsfZ/tZ0P7jAjeWMeiGHe0umu/hbbbSt7X+aejxo0Eetcrrj3x04hXIh8sPAg113lN3jKx5rp2yfB5wnaW/KuIqXA38k6X2Ui3b/NO0BhvdOyojXHUY8dwON3knpCRY7OtD2AwAkfYhyneIAj2AelhYt5HDf4SxUZcraQ4EfeQTzl4zYXTX4VmoCqlywhfEN+pD0VuBpwFrgAa50V6RernczlmHL+w/KgLWPdm9cz6BMuFYr3Jd6wN2wbE9KWlqpjNb8fuqB7Zu6Eeq7ZbB301pMnbEfLGmH18IomuUWcpv72G5NN2qSpp1q1fbzKpXzG0pXu9cBX7HtEbUf30y5mcp2Bs+iV+0NqxWSNts+eLbbbs/G1ZNtHCQtZ5pmOY/grlwLOdw32r5/9/hlwGPcc2u6cfWB3p1Iejml/XhvymCvcym3Wasa7jF7kj5GGXz1wb71z6eMuH3W/NQsxkHl3rav7f/0JmkF8Hrb1ZvlFnK433JFWdJFwCfcze8+zgEuNWj6G/Da9e74NFXeSAd9xOyp3DnoPOBGbr0RzArKBcKnjqhbZCwQMwwuu3zq2kLVMhdwuF9C6XnxI0of8fvZ/nE3GOgK2zXu9jMWKvO532Y13Q14bY/s2sfuOuijVSp3ZJr6J9/oSvPFx8I2H81yC/mC6rhuTTdytl889VhlGshnUy4YX8qO91IcRdm766CPJtm+hHKyErcvGyS9cCfNcjVv6XnrsRfqmft0JL2sf0DQQqfb3oD3b133BrzNDvqI2N3NR7Pc7hru19g+YL7rMSzteAPe0/v7Olcs5wJuHfTxeOBulBfPS3fzQR8RTRhns9zuGu7X2l4y3/UYlsZ3A95bLsyo3Ekqgz4ibqcWcpv7dHa3d6Sx3BqQhgZ9RMSuWbBn7t3EYYMqJ2CvUfYw2V21NOgjInbNgg33lszwRlUtdCXdwfbvZ94zIlqXcG9ICzdjiIg69pjvCkRV45omNyIWuLRbt2Viutt5jeJWXhGxMCXc2zKWW9NFxMKXNveGpM09Iqakzb0tOWOPCCBn7k2RdG/gmcDBwOXA39vePr+1ioj5kHBviKRzKaNUvwwcBfzQ9kvnt1YRMR8S7g3pm1tmMfD1tMFH3D6lzb0tvXPLpDkm4nYsZ+4NydwyETEl4R4R0aA0y0RENCjhHhHRoIR7RESDEu4REQ36/1Ec9oVo/rjYAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "mi.plot.bar()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['CRIM', 'INDUS', 'NOX', 'RM', 'AGE', 'TAX', 'PTRATIO', 'LSTAT'], dtype='object')"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sel = SelectKBest(score_func=mutual_info_regression, k=8)\n",
    "sel.fit(X_train, y_train)\n",
    "X_train.columns[sel.get_support()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_mi = sel.transform(X_train)\n",
    "X_test_mi = sel.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.6946358736819169\n",
      "Wall time: 1.16 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "model = LinearRegression()\n",
    "model.fit(X_train_mi, y_train)\n",
    "y_pred = model.predict(X_test_mi)\n",
    "print(r2_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7606186905933202\n",
      "Wall time: 202 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "model = LinearRegression()\n",
    "model.fit(X_train, y_train)\n",
    "y_pred = model.predict(X_test)\n",
    "print(r2_score(y_test, y_pred))"
   ]
  },
  {
   "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"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}