04_DateNotebook.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature Tools:\n",
    "- This Jupyter Notebook is similar to the 01_DateNotebook, 02_DateNotebook, and 03_DateNotebook with the exception that we are automating feature aggregation with a tool: `Feature Tools`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Importing the necessary packages\n",
    "import numpy as np \n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import featuretools as ft\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Id</th>\n",
       "      <th>v2a1</th>\n",
       "      <th>hacdor</th>\n",
       "      <th>rooms</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>v18q</th>\n",
       "      <th>v18q1</th>\n",
       "      <th>r4h1</th>\n",
       "      <th>...</th>\n",
       "      <th>SQBescolari</th>\n",
       "      <th>SQBage</th>\n",
       "      <th>SQBhogar_total</th>\n",
       "      <th>SQBedjefe</th>\n",
       "      <th>SQBhogar_nin</th>\n",
       "      <th>SQBovercrowding</th>\n",
       "      <th>SQBdependency</th>\n",
       "      <th>SQBmeaned</th>\n",
       "      <th>agesq</th>\n",
       "      <th>Target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ID_279628684</td>\n",
       "      <td>190000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>100</td>\n",
       "      <td>1849</td>\n",
       "      <td>1</td>\n",
       "      <td>100</td>\n",
       "      <td>0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>100.0</td>\n",
       "      <td>1849</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ID_f29eb3ddd</td>\n",
       "      <td>135000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>144</td>\n",
       "      <td>4489</td>\n",
       "      <td>1</td>\n",
       "      <td>144</td>\n",
       "      <td>0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>64.0</td>\n",
       "      <td>144.0</td>\n",
       "      <td>4489</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ID_68de51c94</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>121</td>\n",
       "      <td>8464</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>64.0</td>\n",
       "      <td>121.0</td>\n",
       "      <td>8464</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ID_d671db89c</td>\n",
       "      <td>180000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>81</td>\n",
       "      <td>289</td>\n",
       "      <td>16</td>\n",
       "      <td>121</td>\n",
       "      <td>4</td>\n",
       "      <td>1.777778</td>\n",
       "      <td>1.0</td>\n",
       "      <td>121.0</td>\n",
       "      <td>289</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ID_d56d6f5f5</td>\n",
       "      <td>180000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>121</td>\n",
       "      <td>1369</td>\n",
       "      <td>16</td>\n",
       "      <td>121</td>\n",
       "      <td>4</td>\n",
       "      <td>1.777778</td>\n",
       "      <td>1.0</td>\n",
       "      <td>121.0</td>\n",
       "      <td>1369</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 143 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             Id      v2a1  hacdor  rooms  hacapo  v14a  refrig  v18q  v18q1  \\\n",
       "0  ID_279628684  190000.0       0      3       0     1       1     0    NaN   \n",
       "1  ID_f29eb3ddd  135000.0       0      4       0     1       1     1    1.0   \n",
       "2  ID_68de51c94       NaN       0      8       0     1       1     0    NaN   \n",
       "3  ID_d671db89c  180000.0       0      5       0     1       1     1    1.0   \n",
       "4  ID_d56d6f5f5  180000.0       0      5       0     1       1     1    1.0   \n",
       "\n",
       "   r4h1   ...    SQBescolari  SQBage  SQBhogar_total  SQBedjefe  SQBhogar_nin  \\\n",
       "0     0   ...            100    1849               1        100             0   \n",
       "1     0   ...            144    4489               1        144             0   \n",
       "2     0   ...            121    8464               1          0             0   \n",
       "3     0   ...             81     289              16        121             4   \n",
       "4     0   ...            121    1369              16        121             4   \n",
       "\n",
       "   SQBovercrowding  SQBdependency  SQBmeaned  agesq  Target  \n",
       "0         1.000000            0.0      100.0   1849     4.0  \n",
       "1         1.000000           64.0      144.0   4489     4.0  \n",
       "2         0.250000           64.0      121.0   8464     4.0  \n",
       "3         1.777778            1.0      121.0    289     4.0  \n",
       "4         1.777778            1.0      121.0   1369     4.0  \n",
       "\n",
       "[5 rows x 143 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Storing the datasets\n",
    "train = pd.read_csv('./archive/train.csv')\n",
    "test = pd.read_csv('./archive/test.csv')\n",
    "test['Target'] = np.nan\n",
    "\n",
    "data = pd.concat([train, test])\n",
    "data.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Some households will not have the houseleader with a target variable disregard those, mark them as 4 by default\n",
    "train_valid = train.loc[train['parentesco1'] == 1, ['idhogar', 'Id', 'Target']].copy()\n",
    "test_valid = test.loc[test['parentesco1'] == 1, ['idhogar', 'Id']].copy()\n",
    "\n",
    "submission_base = test[['Id', 'idhogar']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_similar_cols(data, groupby_format):\n",
    "    \"\"\"Returns columns that follow a similar groupby_format\"\"\"\n",
    "    return data.filter(like=groupby_format, axis=1).columns.tolist()\n",
    "\n",
    "def get_corrs_matrix(data, corr_threshold=0.95):\n",
    "    \"\"\" Returns a list: correlation matrix, columns to drop based on the correlation threshold\"\"\"\n",
    "    corr_matrix = data.corr()\n",
    "    \n",
    "    # Select upper triangle of correlation matrix\n",
    "    upper = corr_matrix.where(np.triu(np.ones(corr_matrix.shape), k=1).astype(np.bool))\n",
    "    \n",
    "    # Find index of feature columns with correlation greater than 0.95\n",
    "    to_drop = [c for c in upper.columns if any(abs(upper[c]) > corr_threshold)]\n",
    "    return [corr_matrix, to_drop]\n",
    "\n",
    "def get_dtypes(data, verbose=False):\n",
    "    \"\"\"Returns a list: numerical values, object-like columns\"\"\"\n",
    "    \n",
    "    numerics = ['int16', 'int32', 'int64', 'float16', 'float32', 'float64']\n",
    "    numerics_cols = data.select_dtypes(include=numerics).columns\n",
    "    object_cols = data.select_dtypes(include='object').columns\n",
    "    if verbose:\n",
    "        print('There are {0} numeric cols: {1}\\nThere are {2} object cols: {3}\\nThere are total cols: {4}'.format\n",
    "              (len(numerics_cols),numerics_cols, len(object_cols), object_cols, len(data.columns)))\n",
    "    return [numerics_cols, object_cols]\n",
    "\n",
    "def get_missing_values(data, cols):\n",
    "    \"\"\"Returns a dataframe with missing values (in absolute and percentage format)\"\"\"\n",
    "    \n",
    "    missing_percent = data[cols].apply(lambda x: sum(x.isnull())/len(x), axis=0).sort_values(ascending=False)\n",
    "    missing_abs = data[cols].apply(lambda x: sum(x.isnull()), axis=0).sort_values(ascending=False)\n",
    "    df = pd.DataFrame({'Missing': missing_abs, 'Percent': missing_percent})\n",
    "    \n",
    "    # Removing zero values\n",
    "    df = df.loc[~(df==0).all(axis=1)]\n",
    "    return df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Analysis: dtypes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Id', 'idhogar', 'dependency', 'edjefe', 'edjefa'], dtype='object')"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.select_dtypes('object').columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>dependency</th>\n",
       "      <th>edjefa</th>\n",
       "      <th>edjefe</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>33413.000000</td>\n",
       "      <td>33413.000000</td>\n",
       "      <td>33413.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>1.172238</td>\n",
       "      <td>2.827822</td>\n",
       "      <td>5.170353</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>1.649249</td>\n",
       "      <td>4.606153</td>\n",
       "      <td>5.214174</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>6.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>1.333333</td>\n",
       "      <td>6.000000</td>\n",
       "      <td>9.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>8.000000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         dependency        edjefa        edjefe\n",
       "count  33413.000000  33413.000000  33413.000000\n",
       "mean       1.172238      2.827822      5.170353\n",
       "std        1.649249      4.606153      5.214174\n",
       "min        0.000000      0.000000      0.000000\n",
       "25%        0.333333      0.000000      0.000000\n",
       "50%        0.666667      0.000000      6.000000\n",
       "75%        1.333333      6.000000      9.000000\n",
       "max        8.000000     21.000000     21.000000"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Reshaping the mixed values for object\n",
    "mapping_incorrections = {'yes': 1, 'no': 0}\n",
    "\n",
    "# Reshaping the data based on replace the yes and no\n",
    "for col in ['dependency', 'edjefa', 'edjefe']:\n",
    "    data[col] = data[col].replace(mapping_incorrections).astype(np.float64)\n",
    "    \n",
    "data[['dependency', 'edjefa', 'edjefe']].describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Cleaning: Missing Values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>Missing</th>\n",
       "      <th>Percent</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>rez_esc</th>\n",
       "      <td>27581</td>\n",
       "      <td>0.825457</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v18q1</th>\n",
       "      <td>25468</td>\n",
       "      <td>0.762218</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v2a1</th>\n",
       "      <td>24263</td>\n",
       "      <td>0.726154</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Target</th>\n",
       "      <td>23856</td>\n",
       "      <td>0.713974</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meaneduc</th>\n",
       "      <td>36</td>\n",
       "      <td>0.001077</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SQBmeaned</th>\n",
       "      <td>36</td>\n",
       "      <td>0.001077</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           Missing   Percent\n",
       "rez_esc      27581  0.825457\n",
       "v18q1        25468  0.762218\n",
       "v2a1         24263  0.726154\n",
       "Target       23856  0.713974\n",
       "meaneduc        36  0.001077\n",
       "SQBmeaned       36  0.001077"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_missing_values(data, data.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Filling in the missing value\n",
    "data['v18q1'] = data['v18q1'].fillna(0)\n",
    "\n",
    "# Updating the payment to 0 for people who own the property\n",
    "data.loc[(data['v2a1'].isnull()) & (data['tipovivi1'] == 1), \"v2a1\"] = 0\n",
    "\n",
    "# Creating an additional boolean values of people who have missing values\n",
    "data['v2a1-missing'] = data['v2a1'].isnull()\n",
    "\n",
    "# If individual is over 19 or younger than 7 and missing years behind, set it to 0\n",
    "data.loc[((data['age'] > 19) | (data['age'] < 7)) & (data['rez_esc'].isnull()), 'rez_esc'] = 0\n",
    "\n",
    "# Add a flag for those between 7 and 19 with a missing value\n",
    "data['rez_esc-missing'] = data['rez_esc'].isnull()\n",
    "\n",
    "# Fixing an outlier\n",
    "data.loc[data['rez_esc'] > 5, 'rez_esc'] = 5"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Enginerring: Columns Type"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(33413, 161)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# tamviv: The # of people living in the house, hhsize: The # of people in the household\n",
    "# Thus, the difference could provide some insight\n",
    "data['hhsize-diff'] = data['tamviv'] - data['hhsize']\n",
    "\n",
    "# Correlated Values: coopele\n",
    "elec = []\n",
    "\n",
    "# Assign values for electricity (ordinal values)\n",
    "for i, row in data.iterrows():\n",
    "    if row['noelec'] == 1:\n",
    "        elec.append(0)\n",
    "    elif row['coopele'] == 1:\n",
    "        elec.append(1)\n",
    "    elif row['public'] == 1:\n",
    "        elec.append(2)\n",
    "    elif row['planpri'] == 1:\n",
    "        elec.append(3)\n",
    "    else:\n",
    "        elec.append(np.nan)\n",
    "        \n",
    "# Record the new variable and missing flag\n",
    "data['elec'] = elec\n",
    "data['elec-missing'] = data['elec'].isnull()\n",
    "\n",
    "# Wall Ordinal Variables\n",
    "walls_columns = get_similar_cols(data, groupby_format='epared')\n",
    "data['walls'] = np.argmax(np.array(data[walls_columns]), axis=1)\n",
    "\n",
    "# Roof Ordinal Variables\n",
    "roof_columns = get_similar_cols(data, groupby_format='etecho')\n",
    "data['roof'] = np.argmax(np.array(data[roof_columns]), axis=1)\n",
    "\n",
    "# Floor Ordinal Variables\n",
    "floor_columns = get_similar_cols(data, groupby_format='eviv')\n",
    "data['floor'] = np.argmax(np.array(data[floor_columns]), axis=1)\n",
    "\n",
    "# Per capita features\n",
    "data['phones-per-capita'] = data['qmobilephone'] / data['tamviv']\n",
    "data['tablets-per-capita'] = data['v18q1'] / data['tamviv']\n",
    "data['rooms-per-capita'] = data['rooms'] / data['tamviv']\n",
    "data['rent-per-capita'] = data['v2a1'] / data['tamviv']\n",
    "\n",
    "# No toilet, no electricity, no floor, no water service, no ceiling\n",
    "data['warning'] = 1 * (data['sanitario1'] + (data['elec'] == 0) + \n",
    "                        data['pisonotiene'] + data['abastaguano'] + \n",
    "                        (data['cielorazo'] == 0))\n",
    "\n",
    "# Owns a refrigerator, computer, tablet, and television\n",
    "data['bonus'] = 1 * (data['refrig'] + \n",
    "                     data['computer'] + \n",
    "                     (data['v18q1'] > 0) + \n",
    "                     data['television'])\n",
    "\n",
    "# Create one feature from the `instlevel` columns\n",
    "instl_vars = get_similar_cols(data, 'instl')\n",
    "data['instl_var'] = np.argmax(np.array(data[[c for c in instl_vars]]), axis=1)\n",
    "\n",
    "data['escolari/age'] = data['escolari'] / data['age']\n",
    "data['instl_var/age'] = data['instl_var'] / data['age']\n",
    "data['tech'] = data['v18q'] + data['mobilephone']\n",
    "\n",
    "data.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Cleaning: Removing Columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Removing Squared Columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(33413, 152)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = data[[x for x in data if not x.startswith('SQB')]]\n",
    "data = data.drop(columns = ['agesq'])\n",
    "data.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Removing Correlated Columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['tamhog',\n",
       " 'hhsize',\n",
       " 'coopele',\n",
       " 'female',\n",
       " 'hogar_total',\n",
       " 'area2',\n",
       " 'elec',\n",
       " 'instl_var']"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corr_matrix, to_drop = get_corrs_matrix(data)\n",
    "to_drop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(33413, 144)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Dropping the columns\n",
    "data.drop(to_drop, axis=1, inplace=True)\n",
    "data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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>v2a1</th>\n",
       "      <th>hacdor</th>\n",
       "      <th>rooms</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>v18q</th>\n",
       "      <th>v18q1</th>\n",
       "      <th>r4h1</th>\n",
       "      <th>...</th>\n",
       "      <th>floor</th>\n",
       "      <th>phones-per-capita</th>\n",
       "      <th>tablets-per-capita</th>\n",
       "      <th>rooms-per-capita</th>\n",
       "      <th>rent-per-capita</th>\n",
       "      <th>warning</th>\n",
       "      <th>bonus</th>\n",
       "      <th>escolari/age</th>\n",
       "      <th>instl_var/age</th>\n",
       "      <th>tech</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ID_279628684</td>\n",
       "      <td>190000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>3.00</td>\n",
       "      <td>190000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.232558</td>\n",
       "      <td>0.069767</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ID_f29eb3ddd</td>\n",
       "      <td>135000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>4.00</td>\n",
       "      <td>135000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0.179104</td>\n",
       "      <td>0.104478</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ID_68de51c94</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>8.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.119565</td>\n",
       "      <td>0.043478</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ID_d671db89c</td>\n",
       "      <td>180000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>0.75</td>\n",
       "      <td>0.25</td>\n",
       "      <td>1.25</td>\n",
       "      <td>45000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0.529412</td>\n",
       "      <td>0.176471</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ID_d56d6f5f5</td>\n",
       "      <td>180000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>0.75</td>\n",
       "      <td>0.25</td>\n",
       "      <td>1.25</td>\n",
       "      <td>45000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0.297297</td>\n",
       "      <td>0.108108</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 144 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             Id      v2a1  hacdor  rooms  hacapo  v14a  refrig  v18q  v18q1  \\\n",
       "0  ID_279628684  190000.0       0      3       0     1       1     0    0.0   \n",
       "1  ID_f29eb3ddd  135000.0       0      4       0     1       1     1    1.0   \n",
       "2  ID_68de51c94       0.0       0      8       0     1       1     0    0.0   \n",
       "3  ID_d671db89c  180000.0       0      5       0     1       1     1    1.0   \n",
       "4  ID_d56d6f5f5  180000.0       0      5       0     1       1     1    1.0   \n",
       "\n",
       "   r4h1  ...   floor  phones-per-capita  tablets-per-capita  rooms-per-capita  \\\n",
       "0     0  ...       0               1.00                0.00              3.00   \n",
       "1     0  ...       1               1.00                1.00              4.00   \n",
       "2     0  ...       2               0.00                0.00              8.00   \n",
       "3     0  ...       2               0.75                0.25              1.25   \n",
       "4     0  ...       2               0.75                0.25              1.25   \n",
       "\n",
       "   rent-per-capita  warning  bonus  escolari/age  instl_var/age  tech  \n",
       "0         190000.0        0      1      0.232558       0.069767     1  \n",
       "1         135000.0        0      2      0.179104       0.104478     2  \n",
       "2              0.0        0      1      0.119565       0.043478     0  \n",
       "3          45000.0        0      2      0.529412       0.176471     2  \n",
       "4          45000.0        0      2      0.297297       0.108108     2  \n",
       "\n",
       "[5 rows x 144 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Engineering: Variable Types\n",
    "\n",
    "- We will incorporate `featuretools` to help us automate aggregation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We need to specify the correct variables types:\n",
    "\n",
    "- Individual Variables: these are characteristics of each individual rather than the household\n",
    "    - Boolean: Yes or No (0 or 1)\n",
    "    - Ordered Discrete: Integers with an ordering\n",
    "- Household variables\n",
    "    - Boolean: Yes or No\n",
    "    - Ordered Discrete: Integers with an ordering\n",
    "- Continuous numeric\n",
    "    - Below we manually define the variables in each category. This is a little tedious, but also necessary."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "hh_bool = ['hacdor', 'hacapo', 'v14a', 'refrig', 'paredblolad', 'paredzocalo', \n",
    "           'paredpreb','pisocemento', 'pareddes', 'paredmad',\n",
    "           'paredzinc', 'paredfibras', 'paredother', 'pisomoscer', 'pisoother', \n",
    "           'pisonatur', 'pisonotiene', 'pisomadera',\n",
    "           'techozinc', 'techoentrepiso', 'techocane', 'techootro', 'cielorazo', \n",
    "           'abastaguadentro', 'abastaguafuera', 'abastaguano',\n",
    "            'public', 'planpri', 'noelec', 'coopele', 'sanitario1', \n",
    "           'sanitario2', 'sanitario3', 'sanitario5',   'sanitario6',\n",
    "           'energcocinar1', 'energcocinar2', 'energcocinar3', 'energcocinar4', \n",
    "           'elimbasu1', 'elimbasu2', 'elimbasu3', 'elimbasu4', \n",
    "           'elimbasu5', 'elimbasu6', 'epared1', 'epared2', 'epared3',\n",
    "           'etecho1', 'etecho2', 'etecho3', 'eviv1', 'eviv2', 'eviv3', \n",
    "           'tipovivi1', 'tipovivi2', 'tipovivi3', 'tipovivi4', 'tipovivi5', \n",
    "           'computer', 'television', 'lugar1', 'lugar2', 'lugar3',\n",
    "           'lugar4', 'lugar5', 'lugar6', 'area1', 'area2', 'v2a1-missing', 'elec-missing']\n",
    "\n",
    "hh_ordered = [ 'rooms', 'r4h1', 'r4h2', 'r4h3', 'r4m1','r4m2','r4m3', 'r4t1',  'r4t2', \n",
    "              'r4t3', 'v18q1', 'tamhog','tamviv','hhsize','hogar_nin','hhsize-diff',\n",
    "              'elec',  'walls', 'roof', 'floor', 'walls+roof+floor', 'warning', 'bonus',\n",
    "              'hogar_adul','hogar_mayor','hogar_total',  'bedrooms', 'qmobilephone']\n",
    "\n",
    "hh_cont = ['v2a1', 'dependency', 'edjefe', 'edjefa', 'meaneduc', 'overcrowding',\n",
    "          'phones-per-capita', 'tablets-per-capita', 'rooms-per-capita', 'rent-per-capita']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "ind_bool = ['v18q', 'dis', 'male', 'female', 'estadocivil1', 'estadocivil2', 'estadocivil3', \n",
    "            'estadocivil4', 'estadocivil5', 'estadocivil6', 'estadocivil7', \n",
    "            'parentesco1', 'parentesco2',  'parentesco3', 'parentesco4', 'parentesco5', \n",
    "            'parentesco6', 'parentesco7', 'parentesco8',  'parentesco9', 'parentesco10', \n",
    "            'parentesco11', 'parentesco12', 'instlevel1', 'instlevel2', 'instlevel3', \n",
    "            'instlevel4', 'instlevel5', 'instlevel6', 'instlevel7', 'instlevel8', \n",
    "            'instlevel9', 'mobilephone', 'rez_esc-missing']\n",
    "\n",
    "ind_ordered = ['age', 'escolari', 'rez_esc', 'inst', 'tech']\n",
    "\n",
    "ind_cont = ['escolari/age', 'inst/age']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Removing columns that are not listed in the list (cell above)\n",
    "to_remove = []\n",
    "for l in [hh_ordered, hh_bool, hh_cont, ind_bool, ind_ordered, ind_cont]:\n",
    "    for c in l:\n",
    "        if c not in data:\n",
    "            to_remove.append(c)\n",
    "            \n",
    "for l in [hh_ordered, hh_bool, hh_cont, ind_bool, ind_ordered, ind_cont]:\n",
    "    for c in to_remove:\n",
    "        if c in l:\n",
    "            l.remove(c)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Converting the datatype of the columns in the boolean list to boolean\n",
    "for variable in (hh_bool + ind_bool):\n",
    "    data[variable] = data[variable].astype('bool')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Converting the datatype of the columns in the continous list to floats\n",
    "for variable in (hh_cont + ind_cont):\n",
    "    data[variable] = data[variable].astype(float)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Could not convert rez_esc because of missing values.\n"
     ]
    }
   ],
   "source": [
    "# Converting the datatype of the columns in the ordered list to ints\n",
    "for variable in (hh_ordered + ind_ordered):\n",
    "    try:\n",
    "        data[variable] = data[variable].astype(int)\n",
    "    except Exception as e:\n",
    "        print(f'Could not convert {variable} because of missing values.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "data.dtypes.value_counts().plot.bar(edgecolor = 'k');\n",
    "plt.title('Variable Type Distribution');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes:\n",
    "- The majority of the data measures are booleans."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Tools"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### EntitySet and Entities\n",
    "\n",
    "For a more detailed overview: `/Users/alexguanga/All_Projects/exponential_learning/new_tools`\n",
    "\n",
    "**Entity**: \n",
    "- An entity is simply a table (or a DataFrame if you think in Pandas). \n",
    "\n",
    "**EntitySet**: \n",
    "- A collection of tables and the relationships between them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Entityset: households\n",
       "  Entities:\n",
       "    data [Rows: 33413, Columns: 144]\n",
       "  Relationships:\n",
       "    No relationships"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Creating an instance for households\n",
    "es = ft.EntitySet(id='households')\n",
    "\n",
    "# Building the featuretools archiecture with the entire dataset (entity_id is the id for this framework)\n",
    "es.entity_from_dataframe(entity_id='data', \n",
    "                         dataframe=data, \n",
    "                         index='Id')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Normalize Household Table\n",
    "- **Remember that the data entity is the entire dataset whereas the household entity is at the household level**\n",
    "- Normalization allows us to create another table with the same dataset but separated at another instance. In our example, it's at `idhogar`. \n",
    "    - We are also including additional variables that correspond to the household level!\n",
    "- The new table is derived from the data table and we need to bring along any of the household level variables. \n",
    "- Since these are the same for all members of a household, we can directly add these as columns in the household table using additional_variables. The index of the household table is idhogar which uniquely identifies each household.\n",
    "- `additional_variables` are the variables that will be passed into the household entity\n",
    "- Table Relationships\n",
    "    - Normalizing the entity automatically adds in the relationship between the parent, household, and the child, ind. This relationship links the two tables and allows us to create \"deep features\" by aggregating individuals in each household.\n",
    "    - The best way to think of a relationship between two tables is the analogy of parent to child. This is a one-to-many relationship: each parent can have multiple children. In the realm of tables, a parent table has one row for every child, but the child table may have multiple rows corresponding to multiple children of the same parent."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Entityset: households\n",
       "  Entities:\n",
       "    data [Rows: 33413, Columns: 41]\n",
       "    households [Rows: 10340, Columns: 104]\n",
       "  Relationships:\n",
       "    data.idhogar -> households.idhogar"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "es.normalize_entity(base_entity_id='data', \n",
    "                    new_entity_id='households', \n",
    "                    index='idhogar', \n",
    "                    additional_variables=hh_bool + hh_ordered + hh_cont + ['Target'])\n",
    "es"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Deep Feature Synthesis\n",
    "- Using feature primitives, Deep Feature Synthesis can build hundreds (or 1000s as we will later see) of features from the relationships between tables and the columns in tables themselves. There are two types of primitives, which are operations applied to data:\n",
    "    - Aggregation primitives: \n",
    "        - Operations completed across a parent-to-child (one-to-many) relationship that group by the parent and calculate stats for the children. \n",
    "        - These primitives take related instances as an input and output a single value. They are applied across a parent-child relationship in an entity set. E.g: `\"count\"`, `\"sum\"`, `\"avg_time_between\"`.\n",
    "    - Transform primitives: \n",
    "        - Operations done on a single table to one or more columns (i.e. the difference between two columns in one table).\n",
    "        - These primitives take one or more variables from an entity as an input and output a new variable for that entity. They are applied to a single entity. E.g: `\"hour\"`, `\"time_since_previous\"`, `\"absolute\"`\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!ulimit -n 4096"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Built 173 features\n",
      "EntitySet scattered to workers in 4.376 seconds\n",
      "Elapsed: 01:24 | Remaining: 00:00 | Progress: 100%|██████████| Calculated: 104/104 chunks\n"
     ]
    }
   ],
   "source": [
    "# Deep Feature Synthesis: Aggregation and Transform primitives are build as default\n",
    "# The max_depth parameter controls the maximum depth of the features returned by DFS\n",
    "# With a depth of 2, a number of features are generated using the supplied primitives\n",
    "feature_matrix, feature_names = ft.dfs(entityset=es, \n",
    "                                       target_entity='households', \n",
    "                                       max_depth=2, \n",
    "                                       verbose=1, \n",
    "                                       n_jobs=-1, \n",
    "                                       chunk_size=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "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>hacdor</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>paredblolad</th>\n",
       "      <th>paredzocalo</th>\n",
       "      <th>paredpreb</th>\n",
       "      <th>pisocemento</th>\n",
       "      <th>pareddes</th>\n",
       "      <th>paredmad</th>\n",
       "      <th>...</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel2)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel3)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel4)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel5)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel6)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel7)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel8)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel9)</th>\n",
       "      <th>PERCENT_TRUE(data.mobilephone)</th>\n",
       "      <th>PERCENT_TRUE(data.rez_esc-missing)</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>idhogar</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>000a08204</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>000bce7c4</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001845fb0</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001ff74ca</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>003123ec2</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 173 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           hacdor  hacapo  v14a  refrig  paredblolad  paredzocalo  paredpreb  \\\n",
       "idhogar                                                                        \n",
       "000a08204   False   False  True    True         True        False      False   \n",
       "000bce7c4   False   False  True    True        False         True      False   \n",
       "001845fb0   False   False  True    True         True        False      False   \n",
       "001ff74ca   False   False  True    True         True        False      False   \n",
       "003123ec2   False   False  True    True        False        False       True   \n",
       "\n",
       "           pisocemento  pareddes  paredmad  \\\n",
       "idhogar                                      \n",
       "000a08204        False     False     False   \n",
       "000bce7c4        False     False     False   \n",
       "001845fb0        False     False     False   \n",
       "001ff74ca        False     False     False   \n",
       "003123ec2         True     False     False   \n",
       "\n",
       "                          ...                  PERCENT_TRUE(data.instlevel2)  \\\n",
       "idhogar                   ...                                                  \n",
       "000a08204                 ...                                            0.0   \n",
       "000bce7c4                 ...                                            0.5   \n",
       "001845fb0                 ...                                            0.0   \n",
       "001ff74ca                 ...                                            0.0   \n",
       "003123ec2                 ...                                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel3)  PERCENT_TRUE(data.instlevel4)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                           0.00   \n",
       "000bce7c4                           0.00                           0.00   \n",
       "001845fb0                           0.25                           0.25   \n",
       "001ff74ca                           0.00                           0.00   \n",
       "003123ec2                           0.25                           0.25   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel5)  PERCENT_TRUE(data.instlevel6)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                            0.0   \n",
       "000bce7c4                           0.00                            0.0   \n",
       "001845fb0                           0.25                            0.0   \n",
       "001ff74ca                           0.00                            0.0   \n",
       "003123ec2                           0.00                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel7)  PERCENT_TRUE(data.instlevel8)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                            0.0                       0.666667   \n",
       "000bce7c4                            0.0                       0.000000   \n",
       "001845fb0                            0.0                       0.250000   \n",
       "001ff74ca                            0.0                       0.500000   \n",
       "003123ec2                            0.0                       0.000000   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel9)  PERCENT_TRUE(data.mobilephone)  \\\n",
       "idhogar                                                                    \n",
       "000a08204                            0.0                             1.0   \n",
       "000bce7c4                            0.0                             1.0   \n",
       "001845fb0                            0.0                             1.0   \n",
       "001ff74ca                            0.0                             1.0   \n",
       "003123ec2                            0.0                             1.0   \n",
       "\n",
       "           PERCENT_TRUE(data.rez_esc-missing)  \n",
       "idhogar                                        \n",
       "000a08204                                0.00  \n",
       "000bce7c4                                0.00  \n",
       "001845fb0                                0.25  \n",
       "001ff74ca                                0.00  \n",
       "003123ec2                                0.00  \n",
       "\n",
       "[5 rows x 173 columns]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_features = [str(x.get_name()) for x in feature_names]\n",
    "feature_matrix.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>hacdor</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>paredblolad</th>\n",
       "      <th>paredzocalo</th>\n",
       "      <th>paredpreb</th>\n",
       "      <th>pisocemento</th>\n",
       "      <th>pareddes</th>\n",
       "      <th>paredmad</th>\n",
       "      <th>...</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel2)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel3)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel4)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel5)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel6)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel7)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel8)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel9)</th>\n",
       "      <th>PERCENT_TRUE(data.mobilephone)</th>\n",
       "      <th>PERCENT_TRUE(data.rez_esc-missing)</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>idhogar</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>000a08204</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>000bce7c4</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001845fb0</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001ff74ca</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>003123ec2</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 173 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           hacdor  hacapo  v14a  refrig  paredblolad  paredzocalo  paredpreb  \\\n",
       "idhogar                                                                        \n",
       "000a08204   False   False  True    True         True        False      False   \n",
       "000bce7c4   False   False  True    True        False         True      False   \n",
       "001845fb0   False   False  True    True         True        False      False   \n",
       "001ff74ca   False   False  True    True         True        False      False   \n",
       "003123ec2   False   False  True    True        False        False       True   \n",
       "\n",
       "           pisocemento  pareddes  paredmad  \\\n",
       "idhogar                                      \n",
       "000a08204        False     False     False   \n",
       "000bce7c4        False     False     False   \n",
       "001845fb0        False     False     False   \n",
       "001ff74ca        False     False     False   \n",
       "003123ec2         True     False     False   \n",
       "\n",
       "                          ...                  PERCENT_TRUE(data.instlevel2)  \\\n",
       "idhogar                   ...                                                  \n",
       "000a08204                 ...                                            0.0   \n",
       "000bce7c4                 ...                                            0.5   \n",
       "001845fb0                 ...                                            0.0   \n",
       "001ff74ca                 ...                                            0.0   \n",
       "003123ec2                 ...                                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel3)  PERCENT_TRUE(data.instlevel4)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                           0.00   \n",
       "000bce7c4                           0.00                           0.00   \n",
       "001845fb0                           0.25                           0.25   \n",
       "001ff74ca                           0.00                           0.00   \n",
       "003123ec2                           0.25                           0.25   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel5)  PERCENT_TRUE(data.instlevel6)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                            0.0   \n",
       "000bce7c4                           0.00                            0.0   \n",
       "001845fb0                           0.25                            0.0   \n",
       "001ff74ca                           0.00                            0.0   \n",
       "003123ec2                           0.00                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel7)  PERCENT_TRUE(data.instlevel8)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                            0.0                       0.666667   \n",
       "000bce7c4                            0.0                       0.000000   \n",
       "001845fb0                            0.0                       0.250000   \n",
       "001ff74ca                            0.0                       0.500000   \n",
       "003123ec2                            0.0                       0.000000   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel9)  PERCENT_TRUE(data.mobilephone)  \\\n",
       "idhogar                                                                    \n",
       "000a08204                            0.0                             1.0   \n",
       "000bce7c4                            0.0                             1.0   \n",
       "001845fb0                            0.0                             1.0   \n",
       "001ff74ca                            0.0                             1.0   \n",
       "003123ec2                            0.0                             1.0   \n",
       "\n",
       "           PERCENT_TRUE(data.rez_esc-missing)  \n",
       "idhogar                                        \n",
       "000a08204                                0.00  \n",
       "000bce7c4                                0.00  \n",
       "001845fb0                                0.25  \n",
       "001ff74ca                                0.00  \n",
       "003123ec2                                0.00  \n",
       "\n",
       "[5 rows x 173 columns]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Remove any columns that are a derivations of the target variable \n",
    "# These are created because some of transform primitives might have affected the Target.\n",
    "\n",
    "drop_cols = []\n",
    "for col in feature_matrix:\n",
    "    if col == 'Target':\n",
    "        pass\n",
    "    else:\n",
    "        if 'Target' in col:\n",
    "            drop_cols.append(col)\n",
    "            \n",
    "print(drop_cols)            \n",
    "feature_matrix = feature_matrix[[x for x in feature_matrix if x not in drop_cols]]         \n",
    "feature_matrix.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10340, 173)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_matrix.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['MIN(data.tech)', 'MEAN(data.tech)', 'COUNT(data)']"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Remove features that have an absolute inter-correlation value is greater than 0.99 \n",
    "\n",
    "corr_matrix, to_drop = get_corrs_matrix(feature_matrix, corr_threshold=.99)\n",
    "to_drop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_matrix = feature_matrix[[x for x in feature_matrix if x not in to_drop]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "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>hacdor</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>paredblolad</th>\n",
       "      <th>paredzocalo</th>\n",
       "      <th>paredpreb</th>\n",
       "      <th>pisocemento</th>\n",
       "      <th>pareddes</th>\n",
       "      <th>paredmad</th>\n",
       "      <th>...</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel2)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel3)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel4)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel5)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel6)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel7)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel8)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel9)</th>\n",
       "      <th>PERCENT_TRUE(data.mobilephone)</th>\n",
       "      <th>PERCENT_TRUE(data.rez_esc-missing)</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>idhogar</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>000a08204</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>000bce7c4</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001845fb0</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001ff74ca</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>003123ec2</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 170 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           hacdor  hacapo  v14a  refrig  paredblolad  paredzocalo  paredpreb  \\\n",
       "idhogar                                                                        \n",
       "000a08204   False   False  True    True         True        False      False   \n",
       "000bce7c4   False   False  True    True        False         True      False   \n",
       "001845fb0   False   False  True    True         True        False      False   \n",
       "001ff74ca   False   False  True    True         True        False      False   \n",
       "003123ec2   False   False  True    True        False        False       True   \n",
       "\n",
       "           pisocemento  pareddes  paredmad  \\\n",
       "idhogar                                      \n",
       "000a08204        False     False     False   \n",
       "000bce7c4        False     False     False   \n",
       "001845fb0        False     False     False   \n",
       "001ff74ca        False     False     False   \n",
       "003123ec2         True     False     False   \n",
       "\n",
       "                          ...                  PERCENT_TRUE(data.instlevel2)  \\\n",
       "idhogar                   ...                                                  \n",
       "000a08204                 ...                                            0.0   \n",
       "000bce7c4                 ...                                            0.5   \n",
       "001845fb0                 ...                                            0.0   \n",
       "001ff74ca                 ...                                            0.0   \n",
       "003123ec2                 ...                                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel3)  PERCENT_TRUE(data.instlevel4)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                           0.00   \n",
       "000bce7c4                           0.00                           0.00   \n",
       "001845fb0                           0.25                           0.25   \n",
       "001ff74ca                           0.00                           0.00   \n",
       "003123ec2                           0.25                           0.25   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel5)  PERCENT_TRUE(data.instlevel6)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                            0.0   \n",
       "000bce7c4                           0.00                            0.0   \n",
       "001845fb0                           0.25                            0.0   \n",
       "001ff74ca                           0.00                            0.0   \n",
       "003123ec2                           0.00                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel7)  PERCENT_TRUE(data.instlevel8)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                            0.0                       0.666667   \n",
       "000bce7c4                            0.0                       0.000000   \n",
       "001845fb0                            0.0                       0.250000   \n",
       "001ff74ca                            0.0                       0.500000   \n",
       "003123ec2                            0.0                       0.000000   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel9)  PERCENT_TRUE(data.mobilephone)  \\\n",
       "idhogar                                                                    \n",
       "000a08204                            0.0                             1.0   \n",
       "000bce7c4                            0.0                             1.0   \n",
       "001845fb0                            0.0                             1.0   \n",
       "001ff74ca                            0.0                             1.0   \n",
       "003123ec2                            0.0                             1.0   \n",
       "\n",
       "           PERCENT_TRUE(data.rez_esc-missing)  \n",
       "idhogar                                        \n",
       "000a08204                                0.00  \n",
       "000bce7c4                                0.00  \n",
       "001845fb0                                0.25  \n",
       "001ff74ca                                0.00  \n",
       "003123ec2                                0.00  \n",
       "\n",
       "[5 rows x 170 columns]"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_matrix.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Machine Modeling: Training/Testing Sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "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>idhogar</th>\n",
       "      <th>hacdor</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>paredblolad</th>\n",
       "      <th>paredzocalo</th>\n",
       "      <th>paredpreb</th>\n",
       "      <th>pisocemento</th>\n",
       "      <th>pareddes</th>\n",
       "      <th>...</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel2)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel3)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel4)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel5)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel6)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel7)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel8)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel9)</th>\n",
       "      <th>PERCENT_TRUE(data.mobilephone)</th>\n",
       "      <th>PERCENT_TRUE(data.rez_esc-missing)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>001ff74ca</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>003123ec2</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.250000</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>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>004616164</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>004983866</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.000000</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>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>005905417</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.666667</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>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 171 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     idhogar  hacdor  hacapo  v14a  refrig  paredblolad  paredzocalo  \\\n",
       "0  001ff74ca   False   False  True    True         True        False   \n",
       "1  003123ec2   False   False  True    True        False        False   \n",
       "2  004616164   False   False  True    True        False        False   \n",
       "3  004983866   False   False  True    True        False        False   \n",
       "4  005905417   False   False  True   False        False         True   \n",
       "\n",
       "   paredpreb  pisocemento  pareddes                 ...                  \\\n",
       "0      False        False     False                 ...                   \n",
       "1       True         True     False                 ...                   \n",
       "2      False        False     False                 ...                   \n",
       "3      False         True     False                 ...                   \n",
       "4      False        False     False                 ...                   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel2)  PERCENT_TRUE(data.instlevel3)  \\\n",
       "0                            0.0                           0.00   \n",
       "1                            0.0                           0.25   \n",
       "2                            0.5                           0.00   \n",
       "3                            0.0                           0.00   \n",
       "4                            0.0                           0.00   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel4)  PERCENT_TRUE(data.instlevel5)  \\\n",
       "0                       0.000000                            0.0   \n",
       "1                       0.250000                            0.0   \n",
       "2                       0.000000                            0.5   \n",
       "3                       1.000000                            0.0   \n",
       "4                       0.666667                            0.0   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel6)  PERCENT_TRUE(data.instlevel7)  \\\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",
       "   PERCENT_TRUE(data.instlevel8)  PERCENT_TRUE(data.instlevel9)  \\\n",
       "0                            0.5                            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",
       "   PERCENT_TRUE(data.mobilephone)  PERCENT_TRUE(data.rez_esc-missing)  \n",
       "0                             1.0                                 0.0  \n",
       "1                             1.0                                 0.0  \n",
       "2                             1.0                                 0.0  \n",
       "3                             1.0                                 0.0  \n",
       "4                             1.0                                 0.0  \n",
       "\n",
       "[5 rows x 171 columns]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train = feature_matrix[feature_matrix['Target'].notnull()].reset_index()\n",
    "test = feature_matrix[feature_matrix['Target'].isnull()].reset_index()\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Analysis: Target"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Correlation with Target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "warning                           -0.300071\n",
       "hogar_nin                         -0.268105\n",
       "r4t1                              -0.256750\n",
       "PERCENT_TRUE(data.instlevel1)     -0.249333\n",
       "PERCENT_TRUE(data.instlevel2)     -0.245484\n",
       "overcrowding                      -0.238148\n",
       "PERCENT_TRUE(data.estadocivil1)   -0.215546\n",
       "eviv1                             -0.215123\n",
       "epared1                           -0.212209\n",
       "pisocemento                       -0.209285\n",
       "Name: Target, dtype: float64"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corrs = train.corr()\n",
    "corrs['Target'].sort_values(ascending = True).head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "eviv3                  0.293213\n",
       "cielorazo              0.293420\n",
       "phones-per-capita      0.299235\n",
       "walls                  0.301236\n",
       "MIN(data.escolari)     0.303855\n",
       "floor                  0.305943\n",
       "meaneduc               0.332315\n",
       "MAX(data.escolari)     0.371422\n",
       "MEAN(data.escolari)    0.420514\n",
       "Target                 1.000000\n",
       "Name: Target, dtype: float64"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corrs['Target'].sort_values(ascending=True).dropna().tail(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes:\n",
    "- Based of correlation, we notice that `warning`, `hogar_nin`, and `r4t1` have the worst correlation with `Target`\n",
    "    - `hogar_nin`: Number of children 0 to 19 in household\n",
    "    - `warning`: No toilet, no electricity, no floor, no water service, no ceiling\n",
    "    - `r4t1`: persons younger than 12 years of age\n",
    "- Based of correlation, we notice that `MEAN(data.escolari)`, `MAX(data.escolari)`, and `meaneduc` have the best correlation with `Target`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Subset to Relevant Data\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "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>idhogar</th>\n",
       "      <th>hacdor</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>paredblolad</th>\n",
       "      <th>paredzocalo</th>\n",
       "      <th>paredpreb</th>\n",
       "      <th>pisocemento</th>\n",
       "      <th>pareddes</th>\n",
       "      <th>...</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel2)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel3)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel4)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel5)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel6)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel7)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel8)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel9)</th>\n",
       "      <th>PERCENT_TRUE(data.mobilephone)</th>\n",
       "      <th>PERCENT_TRUE(data.rez_esc-missing)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>001ff74ca</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>003123ec2</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.250000</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>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>004616164</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>004983866</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.000000</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>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>005905417</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.666667</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>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 171 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     idhogar  hacdor  hacapo  v14a  refrig  paredblolad  paredzocalo  \\\n",
       "0  001ff74ca   False   False  True    True         True        False   \n",
       "1  003123ec2   False   False  True    True        False        False   \n",
       "2  004616164   False   False  True    True        False        False   \n",
       "3  004983866   False   False  True    True        False        False   \n",
       "4  005905417   False   False  True   False        False         True   \n",
       "\n",
       "   paredpreb  pisocemento  pareddes                 ...                  \\\n",
       "0      False        False     False                 ...                   \n",
       "1       True         True     False                 ...                   \n",
       "2      False        False     False                 ...                   \n",
       "3      False         True     False                 ...                   \n",
       "4      False        False     False                 ...                   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel2)  PERCENT_TRUE(data.instlevel3)  \\\n",
       "0                            0.0                           0.00   \n",
       "1                            0.0                           0.25   \n",
       "2                            0.5                           0.00   \n",
       "3                            0.0                           0.00   \n",
       "4                            0.0                           0.00   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel4)  PERCENT_TRUE(data.instlevel5)  \\\n",
       "0                       0.000000                            0.0   \n",
       "1                       0.250000                            0.0   \n",
       "2                       0.000000                            0.5   \n",
       "3                       1.000000                            0.0   \n",
       "4                       0.666667                            0.0   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel6)  PERCENT_TRUE(data.instlevel7)  \\\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",
       "   PERCENT_TRUE(data.instlevel8)  PERCENT_TRUE(data.instlevel9)  \\\n",
       "0                            0.5                            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",
       "   PERCENT_TRUE(data.mobilephone)  PERCENT_TRUE(data.rez_esc-missing)  \n",
       "0                             1.0                                 0.0  \n",
       "1                             1.0                                 0.0  \n",
       "2                             1.0                                 0.0  \n",
       "3                             1.0                                 0.0  \n",
       "4                             1.0                                 0.0  \n",
       "\n",
       "[5 rows x 171 columns]"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Removing rows from training set we do not have idhogar\n",
    "train = train[train['idhogar'].isin(list(train_valid['idhogar']))]\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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>idhogar</th>\n",
       "      <th>hacdor</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>paredblolad</th>\n",
       "      <th>paredzocalo</th>\n",
       "      <th>paredpreb</th>\n",
       "      <th>pisocemento</th>\n",
       "      <th>pareddes</th>\n",
       "      <th>...</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel2)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel3)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel4)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel5)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel6)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel7)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel8)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel9)</th>\n",
       "      <th>PERCENT_TRUE(data.mobilephone)</th>\n",
       "      <th>PERCENT_TRUE(data.rez_esc-missing)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>000a08204</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>000bce7c4</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>001845fb0</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>003514e22</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>003b51a87</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>...</td>\n",
       "      <td>0.75</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 171 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     idhogar  hacdor  hacapo  v14a  refrig  paredblolad  paredzocalo  \\\n",
       "0  000a08204   False   False  True    True         True        False   \n",
       "1  000bce7c4   False   False  True    True        False         True   \n",
       "2  001845fb0   False   False  True    True         True        False   \n",
       "3  003514e22   False   False  True    True        False        False   \n",
       "4  003b51a87   False   False  True    True        False        False   \n",
       "\n",
       "   paredpreb  pisocemento  pareddes                 ...                  \\\n",
       "0      False        False     False                 ...                   \n",
       "1      False        False     False                 ...                   \n",
       "2      False        False     False                 ...                   \n",
       "3       True        False     False                 ...                   \n",
       "4      False        False      True                 ...                   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel2)  PERCENT_TRUE(data.instlevel3)  \\\n",
       "0                           0.00                           0.00   \n",
       "1                           0.50                           0.00   \n",
       "2                           0.00                           0.25   \n",
       "3                           0.00                           0.25   \n",
       "4                           0.75                           0.25   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel4)  PERCENT_TRUE(data.instlevel5)  \\\n",
       "0                           0.00                           0.00   \n",
       "1                           0.00                           0.00   \n",
       "2                           0.25                           0.25   \n",
       "3                           0.25                           0.00   \n",
       "4                           0.00                           0.00   \n",
       "\n",
       "   PERCENT_TRUE(data.instlevel6)  PERCENT_TRUE(data.instlevel7)  \\\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",
       "   PERCENT_TRUE(data.instlevel8)  PERCENT_TRUE(data.instlevel9)  \\\n",
       "0                       0.666667                            0.0   \n",
       "1                       0.000000                            0.0   \n",
       "2                       0.250000                            0.0   \n",
       "3                       0.250000                            0.0   \n",
       "4                       0.000000                            0.0   \n",
       "\n",
       "   PERCENT_TRUE(data.mobilephone)  PERCENT_TRUE(data.rez_esc-missing)  \n",
       "0                             1.0                                0.00  \n",
       "1                             1.0                                0.00  \n",
       "2                             1.0                                0.25  \n",
       "3                             1.0                                0.00  \n",
       "4                             1.0                                0.00  \n",
       "\n",
       "[5 rows x 171 columns]"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Removing rows from testing set where we do not have idhogar\n",
    "test = test[test['idhogar'].isin(list(test_valid['idhogar']))]\n",
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Labels for training\n",
    "train_labels = np.array(train.pop('Target')).reshape(-1, )\n",
    "test_ids = list(test.pop('idhogar'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes:\n",
    "- Align is a pandas function that returns both dataframes by aligning the column names\n",
    "- https://stackoverflow.com/questions/51645195/pandas-align-function-illustrative-example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2973, 169)"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train, test = train.align(test, axis = 1, join = 'inner')\n",
    "all_features = list(train.columns)\n",
    "train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2973, 169)"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7334, 169)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Modeling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "\n",
    "# Visualization\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from IPython.display import display\n",
    "\n",
    "from collections import Counter\n",
    "from sklearn.metrics import f1_score, make_scorer\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "\n",
    "import lightgbm as lgb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "def macro_f1_score(labels, predictions):\n",
    "    # Reshape the predictions as needed\n",
    "    predictions = predictions.reshape(len(np.unique(labels)), -1).argmax(axis = 0)\n",
    "    \n",
    "    metric_value = f1_score(labels, predictions, average = 'macro')\n",
    "    \n",
    "    # Return is name, value, is_higher_better\n",
    "    return 'macro_f1', metric_value, True"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes:\n",
    "- *The F1 score can be interpreted as a weighted average of the precision and recall, where an F1 score reaches its best value at 1 and worst score at 0. The relative contribution of precision and recall to the F1 score are equal. The formula for the F1 score is:*\n",
    "    - `F1 = 2 * (precision * recall) / (precision + recall)`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model_gbm(features, labels, test_features, test_ids, \n",
    "              nfolds = 5, return_preds = False, hyp = None):\n",
    "    \"\"\"Model using the GBM and cross validation.\n",
    "       Trains with early stopping on each fold.\n",
    "       Hyperparameters probably need to be tuned.\"\"\"\n",
    "    \n",
    "    feature_names = list(features.columns)\n",
    "    \n",
    "    hyp_OPTaaS = { \n",
    "        'boosting_type': 'dart',\n",
    "        'colsample_bytree': 0.9843467236959204,\n",
    "        'learning_rate': 0.11598629586769524,\n",
    "        'min_child_samples': 44,\n",
    "        'num_leaves': 49,\n",
    "        'reg_alpha': 0.35397370408131534,\n",
    "        'reg_lambda': 0.5904910774606467,\n",
    "        'subsample': 0.6299872254632797,\n",
    "        'subsample_for_bin': 60611\n",
    "    }\n",
    "\n",
    "    model = lgb.LGBMClassifier(**hyp_OPTaaS, class_weight = 'balanced',\n",
    "                               objective = 'multiclass', n_jobs = -1, n_estimators = 10000)\n",
    "    \n",
    "    # Using stratified kfold cross validation\n",
    "    strkfold = StratifiedKFold(n_splits = nfolds, shuffle = True)\n",
    "    predictions = pd.DataFrame()\n",
    "    importances = np.zeros(len(feature_names))\n",
    "    \n",
    "    # Convert to arrays for indexing\n",
    "    features = np.array(features)\n",
    "    test_features = np.array(test_features)\n",
    "    labels = np.array(labels).reshape((-1 ))\n",
    "    \n",
    "    valid_scores = []\n",
    "    \n",
    "    # Iterate through the folds\n",
    "    for i, (train_indices, valid_indices) in enumerate(strkfold.split(features, labels)):\n",
    "        # Dataframe for the predictions of each fold\n",
    "        fold_predictions = pd.DataFrame()\n",
    "        \n",
    "        # Training and validation data\n",
    "        X_train = features[train_indices]\n",
    "        X_valid = features[valid_indices]\n",
    "        y_train = labels[train_indices]\n",
    "        y_valid = labels[valid_indices]\n",
    "        \n",
    "        # Train with early stopping\n",
    "        model.fit(X_train, y_train, early_stopping_rounds = 100, \n",
    "                  eval_metric = macro_f1_score,\n",
    "                  eval_set = [(X_train, y_train), (X_valid, y_valid)],\n",
    "                  eval_names = ['train', 'valid'],\n",
    "                  verbose = 200)\n",
    "        \n",
    "        # Record the validation fold score\n",
    "        valid_scores.append(model.best_score_['valid']['macro_f1'])\n",
    "        \n",
    "        # Make predictions from the fold\n",
    "        fold_probabilitites = model.predict_proba(test_features)\n",
    "        \n",
    "        # Record each prediction for each class as a column\n",
    "        for j in range(4):\n",
    "            fold_predictions[(j + 1)] = fold_probabilitites[:, j]\n",
    "            \n",
    "        fold_predictions['idhogar'] = test_ids\n",
    "        fold_predictions['fold'] = (i+1)\n",
    "        predictions = predictions.append(fold_predictions)\n",
    "        \n",
    "        importances += model.feature_importances_ / nfolds   \n",
    "        \n",
    "        display(f'Fold {i + 1}, Validation Score: {round(valid_scores[i], 5)}, Estimators Trained: {model.best_iteration_}')\n",
    "\n",
    "    feature_importances = pd.DataFrame({'feature': feature_names,\n",
    "                                        'importance': importances})\n",
    "    valid_scores = np.array(valid_scores)\n",
    "    display(f'{nfolds} cross validation score: {round(valid_scores.mean(), 5)} with std: {round(valid_scores.std(), 5)}.')\n",
    "    \n",
    "    # If we want to examine predictions don't average over folds\n",
    "    if return_preds:\n",
    "        predictions['Target'] = predictions[[1, 2, 3, 4]].idxmax(axis = 1)\n",
    "        predictions['confidence'] = predictions[[1, 2, 3, 4]].max(axis = 1)\n",
    "        return predictions, feature_importances\n",
    "    \n",
    "    # Average the predictions over folds\n",
    "    predictions = predictions.groupby('idhogar', as_index = False).mean()\n",
    "    \n",
    "    # Find the class and associated probability\n",
    "    predictions['Target'] = predictions[[1, 2, 3, 4]].idxmax(axis = 1)\n",
    "    predictions['confidence'] = predictions[[1, 2, 3, 4]].max(axis = 1)\n",
    "    predictions = predictions.drop(columns = ['fold'])\n",
    "    \n",
    "    # Merge with the base to have one prediction for each individual\n",
    "    submission = submission_base.merge(predictions[['idhogar', 'Target']], on = 'idhogar', how = 'left').drop(columns = ['idhogar'])\n",
    "        \n",
    "    submission['Target'] = submission['Target'].fillna(4).astype(np.int8)\n",
    "    \n",
    "    # return the submission and feature importances\n",
    "    return submission, feature_importances, valid_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# We need to make sure the length of the labels matches the length of the training dataset.\n",
    "len(train_labels) == train.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Verify that the len of test_ids is the same as the length of the testing dataset\n",
    "len(test_ids) == test.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Preventing an issue that occurs with Macs & Gradient Boosting Libraries\n",
    "import os\n",
    "\n",
    "os.environ['KMP_DUPLICATE_LIB_OK']='True'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Fold 1, Validation Score: 0.433, Estimators Trained: 47'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 2, Validation Score: 0.45327, Estimators Trained: 157'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 3, Validation Score: 0.39021, Estimators Trained: 82'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 4, Validation Score: 0.39591, Estimators Trained: 61'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 5, Validation Score: 0.44124, Estimators Trained: 94'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'5 cross validation score: 0.42273 with std: 0.02513.'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%capture --no-display\n",
    "submission, feature_importances, valid_scores = model_gbm(train, train_labels, test, test_ids, 5)\n",
    "\n",
    "results = pd.DataFrame({'version': ['default_5fold'], \n",
    "                        'F1-mean': [valid_scores.mean()], \n",
    "                        'F1-std': [valid_scores.std()]})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Importances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_feature_importances(df, n = 15, return_features = False, threshold = None):\n",
    "    # Sort features with most important at the head\n",
    "    df = df.sort_values('importance', ascending = False).reset_index(drop = True)\n",
    "    \n",
    "    # Normalize the feature importances to add up to one and calculate cumulative importance\n",
    "    df['importance_normalized'] = df['importance'] / df['importance'].sum()\n",
    "    df['cumulative_importance'] = np.cumsum(df['importance_normalized'])\n",
    "    \n",
    "    plt.rcParams['font.size'] = 12\n",
    "    plt.style.use('fivethirtyeight')\n",
    "    # Bar plot of n most important features\n",
    "    df.loc[:n, :].plot.barh(y = 'importance_normalized', \n",
    "                            x = 'feature', color = 'blue', \n",
    "                            edgecolor = 'k', figsize = (12, 5),\n",
    "                            legend = False, linewidth = 2)\n",
    "\n",
    "    plt.xlabel('Normalized Importance', size = 18); plt.ylabel(''); \n",
    "    plt.title(f'Top {n} Most Important Features', size = 18)\n",
    "    plt.gca().invert_yaxis()\n",
    "    \n",
    "    \n",
    "    if threshold:\n",
    "        # Cumulative importance plot\n",
    "        plt.figure(figsize = (12, 5))\n",
    "        plt.plot(list(range(len(df))), df['cumulative_importance'], 'b-')\n",
    "        plt.xlabel('Number of Features', size = 16); plt.ylabel('Cumulative Importance', size = 16); \n",
    "        plt.title('Cumulative Feature Importance', size = 18);\n",
    "        \n",
    "        # Number of features needed for threshold cumulative importance\n",
    "        # This is the index (will need to add 1 for the actual number)\n",
    "        importance_index = np.min(np.where(df['cumulative_importance'] > threshold))\n",
    "        \n",
    "        # Add vertical line to plot\n",
    "        plt.vlines(importance_index + 1, ymin = 0, ymax = 1.05, linestyles = '--', colors = 'red')\n",
    "        plt.show();\n",
    "        \n",
    "        print('{} features required for {:.0f}% of cumulative importance.'.format(importance_index + 1, \n",
    "                                                                                  100 * threshold))\n",
    "    if return_features:\n",
    "        return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_feature_importances(feature_importances)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "#submission.to_csv('ft_baseline.csv', index = False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbUAAAESCAYAAAB6s0uLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzt3X9YVVW+x/E3ohVqiuEBNEOnJATH0ixAJE2wFKlRjNTqpsNkqGSjFIplZqWTYKaYP8jxR7/GapIs8VaWPmCBEse5ajRiilPij0sQ5FEwMYVz//BhX4+gqR07sv28nqcn2eu711l7HfTD2nuzj5vNZrMjIiJiAk1cPQARERFnUaiJiIhpKNRERMQ0FGoiImIaCjURETENhZqIiJiGQu0Kd9111xEeHk5oaCi9e/dm4cKF1NbWArBt2zYmT5581n2Li4tZtWrVWdtLSkoYOXIkACtXrmTSpEkXNLaVK1dSUlJifP3EE0/w7bffXlAfF+L48eMMHjyY8PBwVq9e7dA2btw4brnlFsLDw+nTpw9Wq/U3vVa3bt2oqKgA4J577jln7ZnzcD6Ki4vp1avXeW8/m3HjxrFmzZrf/Lpn89FHHxEcHMy999573vuc6frrr7+g+pycHIYPHw7AJ598wrx5885Z/7e//Y2NGzees5+Lcfr3gDhPU1cPQFzLw8OD3NxcAH788UdGjx7NkSNHeOaZZ+jRowc9evQ467779u0jIyODBx54oF7byZMnadeuHW+99dZFj+2dd94hKCiIdu3aAbBgwYKL7ut8FBQUcOLECWM+zjRjxgwGDx5MVlYWEydOZPPmzQ7tJ0+epGnTC/8r9fnnn5+z/cx5MJO3336bOXPm0KdPn/Oqv9g5PptBgwYxaNCgc9ZMnTrVaa8nl55WamKwWCzMnz+fpUuXYrfbHX4Szc3NJTw8nPDwcO68804qKyt54YUXyMvLIzw8nEWLFrFy5UpGjRrF8OHDiYmJqfdT+4EDB7j//vu5/fbbSUlJAer/ZL9gwQJmzZrFmjVr2L59O4899hjh4eEcO3aM6Ohotm3bBkBGRgZhYWH06tWL6dOnG/tff/31zJgxg969e9O/f3/KysrqHeehQ4d46KGHCAsLo3///vz73//mxx9/JD4+nn//+9+Eh4fz/fffn3WewsLCjPbo6GhefPFFBg0aRHp6OuXl5TzyyCP069ePfv368dVXXwHw008/ERMTw5133snEiROx2///mQenrzTmz59PWFgYvXv35vnnn29wHrZv386gQYPo27cvQ4cO5YcffgBg+/bt9O7dm7vvvptly5ad57t+yptvvkm/fv3o3bs3jzzyCD///LPRtnHjRqKioujZsyfr1q0DoKamhmnTptGvXz/CwsJ4/fXX6/W5c+dOIiIiCA8PJywsjP/85z8O7ampqXz11Vc8+eSTTJs2jerqahISEggLC+POO+/kyy+/BKj3fXU2OTk5REdHM3LkSO644w4ee+wxY543bNjAHXfcwcCBA1m7dq2xT90ZhMOHD9OtWzfjLMXPP/9M165dOXHihMNq9Wz9zJo1y+GHrl69elFcXAzAQw89RN++fQkNDeWNN96oN+6jR48ybNgwevfuTa9eveqdJZALo1ATB506daK2tpYff/zRYfuCBQuYM2cOubm5fPrpp3h4eDB9+nR69epFbm4ujz/+OABbtmzhtddec/gLX2fr1q0sXbqUnJwc1qxZYwRUQwYPHkz37t1ZunQpubm5eHh4GG0lJSU8//zzZGZmkpOTw9atW/nv//5v4NQ/ELfffjubNm0iLCyMN998s17fL730ErfccgubN29m2rRpjB07FovFwquvvmoczx/+8Iezjm3dunUEBQUZXx8+fJhPPvmEJ554gilTppCQkEB2djZvvfUWf/3rXwFISUkhNDSUnJwcoqKiOHDgQL1+169fz8cff8yGDRvYtGkTEyZMqDcPTZs2ZfLkybz11lt88cUX/Nd//RczZswAICEhgdTUVNavX3/WsZ/NfffdR3Z2Nps2bSIgIIC3337baNu3bx8ff/wx77//Pk8++STV1dW8/fbbtGrViuzsbLKzs3nzzTfZu3evQ58rVqxg7Nix5ObmsnHjRtq3b+/QnpycbBzbjBkzWLp0KQCbN29m+fLlJCQkUF1dDZz7++p033zzDbNmzSI/P5+9e/fy1VdfUV1dzYQJE3jvvff49NNPG/xBp3Xr1vzxj380Vunr1q0jIiKCZs2aGTXn009DFi1axBdffEF2djZLlizhp59+cmjfsGEDvr6+bNq0iby8PCIjI8+rX2mYQk3qOX0VUSc0NJSpU6fy2muvcfjw4bOeArrrrrto06bNWduuu+46PDw8uPfee8nLy7uo8W3dupXevXvTtm1bmjZtygMPPGCcCrzqqqsYOHAgAN27d2ffvn319v/qq68YMWIEAH379uXQoUMcPnz4V1932rRphIeH88Ybbzj8VH766mHjxo1MmjSJ8PBwHnzwQSorK6msrGTz5s3GqnfAgAF4enrW63/jxo08/PDDNG/eHKDBeSwqKmLnzp0MGTKE8PBw5syZw//+7/9y+PBhjhw5Qnh4OMAFX+spLCwkKiqKsLAw3n//fYdrl0OGDKFJkybcdNNNdOzYkd27d5OVlcV7771HeHg4kZGR/PTTT3z33XcOfQYHBzN37lzS0tLYv3+/ww8mDfnqq6+Mcd98883ccMMN7NmzBzj399XpbrvtNq6//nqaNGlCt27d2LdvH7t378bPz4+bbroJNzc3hg0b1uC+Q4cO5cMPPwTggw8+YOjQoQ7t59vPmV577TXjzMHBgwfrrVi7du3Kxo0bmT59Ops3b6Z169bn1a80TNfUxMHevXtxd3fHYrGwa9cuY3tiYiL33HMP69ev5+677+ajjz5qcP8WLVqctW83N7d6Xzdt2tQ45QMYP5mfS0OhW6dZs2bG67i7u3Py5Mnz2v/MsTWk7pramU4/5traWtavX/+r/4A3xG63/+o47HY7Xbp0qbcas9lsF/x6p0tISGDlypV069aNlStXOlxXbOh9s9vtzJ49u96qou6UG8ADDzzA7bffzmeffcbQoUN59dVX6du371nHcK739VzfV6e7+uqrjT+f/v6fz/sbFRXFCy+8wKFDh/j6668bvM53tn7O9n2ck5PDF198wfr162nevDnR0dH1vsc7d+7MF198weeff86LL75Iv379SE5O/vWDlQZppSaG8vJyEhMTeeyxx+r95f3+++/p2rUrEydOpHv37uzevZtrr72WysrK8+5/48aNHDp0iGPHjvHxxx8TGhqKt7c3P/74Iz/99BPHjx/ns88+M+pbtmzZYP91pxcrKiqoqanhgw8+oHfv3uc9jrrVCJz6R+e6666jVatW573/uURERBin0eDUzSd1r1l3p+j69esbDKGIiAj+8Y9/GNezDh06BDjOg7+/P+Xl5cbdlydOnGDnzp14enrSqlUrY/V7rrtSG1JVVYWvry8nTpyot++aNWuora3l+++/p7i4GH9/fyIjI1m+fDknTpwAYM+ePRw9etRhv71799KpUyfGjh1LVFQUO3bsOOcYTp+jPXv2sH//fvz9/S/oOBpy8803s2/fPuM66AcffNBgXcuWLenZsyfJyckMGDAAd3f38+7Hz8+Pr7/+Gjh1bbMu3I8cOULr1q1p3rw5u3fv5l//+le91y0pKcHDw4Phw4czfvx4ox+5OFqpXeGOHTtGeHg4J0+exN3dnREjRhjXx063ePFicnNzadKkCV26dOHuu++mSZMmNG3alN69e/PQQw81eErtdKGhoYwZM4bvvvuO2NhY487KyZMnExkZSceOHR3+EXvooYd48sknueaaaxxWJr6+vkyfPp377rsPu93O3XffTXR09Hkf89NPP23ckNC8eXPS09PPe99fk5qaSlJSEmFhYdTU1BAWFsa8efOYMmUKjz76KH369KF379506NCh3r79+/fnm2++oV+/fjRr1ox77rmH5557rt48vPnmmyQnJ3PkyBFqamoYN24cgYGBLF68mMcff5zmzZsTERFx1jEWFRU5XBN86aWXmDp1KpGRkdxwww0EBQVRVVVltHfu3Jno6GjKysqYO3cu11xzDSNHjmTfvn307dsXu92Ol5cXK1eudHid1atX8/7779O0aVN8fHx+dfUxevRoEhMTCQsLw93dncWLFzusvC7WNddcQ1paGsOGDcPLy4vQ0FB27tzZYO3QoUMZNWqUcY32fPv505/+ZJyOve222+jcuTNw6j19/fXXCQsLw9/fn9tvv71ev4WFhUybNo0mTZrQrFkz5s6d+5uP+Urmpo+eERERs9DpRxERMQ2FmoiImIZCTURETEOhJiIipqFQExER0zivUNu0aRMjRowgMDAQT0/Perfunm7ChAl4enrWe/js8ePHmTRpEjfeeCPt27dnxIgRHDx40KFm//79DB8+nPbt23PjjTcyefJkfvnlF4ea3Nxc+vbti4+PD7feeisrVqw432MVERGTO69QO3r0KEFBQaSkpJzzSQlr1qxh69atDT5N/Omnn2bt2rUsX76cTz75hMrKSoYPH05NTQ1w6gGpw4cPp6qqik8++YTly5eTmZnp8ITsvXv3MmzYMIKDg/nyyy958sknmTx58gV9NIaIiJjXeYVa3S+BDh48mCZNGt5l3759TJkyhWXLltV7LuDhw4d5++23jUfAdO/enSVLlrBjxw7jc4qysrLYuXMnS5YsoXv37vTr148XXniBt956iyNHjgDw+uuv4+vry8svv0xAQACjRo3iwQcfZOHChb9hCi5PRUVFrh7CFUnz7hqad9cw47w75ZrayZMnGT16NElJSQQEBNRr3759OydOnHB4ykGHDh0ICAggPz8fAKvVSkBAgMOTFiIjIzl+/Djbt283as58UkJkZCTbtm0zHtcjIiJXLqc8JmvWrFm0adOGRx99tMH2srIy3N3d8fLycthusViMj28oKyvDYrE4tHt5eeHu7u5Qc9ddd9Xr4+TJk1RUVODr69vg6zfWn0Ya67gbO827a2jeXaMxzvu5ngn6m0MtNzeXd955h5ycnAve98ynkp/tCdjnqql7sve5nsLtjIei/t6Kiooa5bgbO827a2jeXcOM8/6bTz/m5OTwww8/EBAQgJeXF15eXuzfv5/p06cbD0319vampqaGiooKh33Ly8uN1Zm3t3e9D92rewr7uWrKy8tp2rQp11133W89FBERaeR+c6iNHj2aTZs2kZOTY/zXrl07EhISjLsSu3fvTrNmzcjOzjb2O3jwILt27SIkJAQ49YGCu3btcrjNPzs7m6uvvpru3bsbNXU3lpxe06NHD4dPqBURkSvTeZ1+rKqqMj7Vtra2lgMHDlBQUECbNm244YYb6l0Lq/uoibplbevWrXnkkUd47rnnsFgstGnThqlTp9K1a1fjGllERASBgYGMHTuWmTNncujQIZ577jlGjhxpfNZVXFwcS5cuZcqUKcTFxZGfn88777zDsmXLnDUfIiLSiJ3XSm3btm306dOHPn36cOzYMWbNmkWfPn146aWXzvuFXnrpJe69917i4uIYOHAgLVq04L333jM+iM/d3Z1//vOfNG/enIEDBxIXF8e9997LzJkzjT46derE+++/z+bNm7nzzjuZM2cOqampDX4asYiIXHn0eWqXKTNewG0MNO+uoXl3DTPOuz75WkTkMuXp2foSv0L9T+J2Fpvt8CXr+1z0QGMRETENhZqIiJiGQk1ERExDoSYiIqahUBMREdNQqImIiGko1ERExDQUaiIiYhoKNRERMQ2FmoiImIZCTURETEOhJiIipqFQExER01CoiYiIaSjURETENBRqIiJiGgo1ERExDYWaiIiYhkJNRERMQ6EmIiKmoVATERHTOK9Q27RpEyNGjCAwMBBPT09WrlxptJ04cYLp06cTFhZG+/btCQgIYPTo0ezfv9+hj+PHjzNp0iRuvPFG2rdvz4gRIzh48KBDzf79+xk+fDjt27fnxhtvZPLkyfzyyy8ONbm5ufTt2xcfHx9uvfVWVqxYcbHHLiIiJnNeoXb06FGCgoJISUnBw8PDoe3nn3/m66+/JikpiS+++IJ33nmHgwcPEhsby8mTJ426p59+mrVr17J8+XI++eQTKisrGT58ODU1NQDU1NQwfPhwqqqq+OSTT1i+fDmZmZlMnTrV6GPv3r0MGzaM4OBgvvzyS5588kkmT57MmjVrnDEXIiLSyLnZbDb7hexw/fXXM3v2bB5++OGz1nz77beEhoayadMmunbtyuHDh+ncuTOLFi1i2LBhABw4cIBu3bqRkZFBZGQk69evZ9iwYXzzzTd06NABgH/+85/89a9/paioiFatWjF9+nTWrl3L1q1bjdd64okn+Pbbb1m/fv3FHP9lq6ioCH9/f1cP44qjeXcNzXvDPD1bu3oIF81mO+yS170k19QqKysB8PT0BGD79u2cOHGCiIgIo6ZDhw4EBASQn58PgNVqJSAgwAg0gMjISI4fP8727duNmtP7qKvZtm0bJ06cuBSHIiIijUhTZ3f4yy+/8OyzzzJw4ECuv/56AMrKynB3d8fLy8uh1mKxUFZWZtRYLBaHdi8vL9zd3R1q7rrrrnp9nDx5koqKCnx9fRscU1FRkTMO7XfXWMfd2GneXUPz3pDbXT2Ai3Yp389zreqdGmonT54kPj6ew4cP8+677/5qvd1ux83Nzfj69D+f7lw1drv9nPvCuSfgcqXTMa6heXcNzbv5uOr9dNrpx5MnT/Loo4+yY8cO1qxZw3XXXWe0eXt7U1NTQ0VFhcM+5eXlxurM29vbWJHVqaiooKam5pw15eXlNG3a1OH1RETkyuSUUDtx4gRxcXHs2LGDtWvX4uPj49DevXt3mjVrRnZ2trHt4MGD7Nq1i5CQEACCg4PZtWuXw23+2dnZXH311XTv3t2o2bhxo0Pf2dnZ9OjRg2bNmjnjUEREpBE7r9OPVVVVfPfddwDU1tZy4MABCgoKaNOmDe3atWPUqFFs27aNd999Fzc3N0pLSwFo1aoVHh4etG7dmkceeYTnnnsOi8VCmzZtmDp1Kl27djWukUVERBAYGMjYsWOZOXMmhw4d4rnnnmPkyJG0atUKgLi4OJYuXcqUKVOIi4sjPz+fd955h2XLll2CqRERkcbmvG7pz8nJ4b777qu3/cEHH2TKlCnceuutDe63aNEi49b/6upqpk2bRkZGBtXV1fTp04dXXnnF4W7H/fv3k5SUxJdffsk111xDbGwsM2fO5OqrrzZqcnNzeeaZZ/j222/x9fVl4sSJ/OUvf7ngA7/c6RqDa2jeXUPz3jDd0n/hLvj31OT3ob/krqF5dw3Ne8MUahdOz34UERHTUKiJiIhpKNRERMQ0FGoiImIaCjURETENhZqIiJiGQk1ERExDoSYiIqahUBMREdNQqImIiGko1ERExDQUaiIiYhoKNRERMQ2FmoiImIZCTURETEOhJiIipqFQExER01CoiYiIaSjURETENBRqIiJiGgo1ERExjfMKtU2bNjFixAgCAwPx9PRk5cqVDu12u51Zs2bRpUsXfH19iY6OZufOnQ41NpuN+Ph4/Pz88PPzIz4+HpvN5lCzY8cOBg0ahK+vL4GBgaSmpmK32x1q1qxZQ0hICN7e3oSEhLB27dqLOW4RETGh8wq1o0ePEhQUREpKCh4eHvXa58+fz6JFi0hNTSUrKwuLxUJMTAyVlZVGzejRoykoKGDVqlVkZGRQUFDAmDFjjPYjR44QExODt7c3WVlZpKSksGDBAhYuXGjUWK1W/vKXv/DAAw+Qk5PDAw88wJ///Gf+9a9//ZY5EBERk3Cz2Wz2Xy/7f9dffz2zZ8/m4YcfBk6t0rp06cJjjz1GUlISAMeOHcPf358ZM2YQFxfHrl27CAkJYd26dYSGhgKQl5dHVFQUW7Zswd/fn+XLl/P888+ze/duIzhffvllVqxYQWFhIW5ubsTFxXHo0CE++ugjYzyDBw+mbdu2LF++3CkTcrkoKirC39/f1cO44mjeXUPz3jBPz9auHsJFs9kOu+R1f/M1teLiYkpLS4mIiDC2eXh4EBYWRn5+PnBqhdWyZUtCQkKMmtDQUFq0aOFQ06tXL4eVYGRkJCUlJRQXFwOwZcsWh9epq6nrQ0RErmy/OdRKS0sBsFgsDtstFgtlZWUAlJWV4eXlhZubm9Hu5uZG27ZtHWoa6qOure61zvU6IiJyZWvqrI5ODyw4dVryzBA706/V1N0k8ms1DfV9uqKiol8Z/eWpsY67sdO8u4bmvSG3u3oAF+1Svp/nOlX9m0PNx8cHOLWa6tChg7G9vLzcWFV5e3tTXl7uEEB2u52KigqHmjNXXOXl5cD/r9h8fHwarDlz9XamxniuXtcYXEPz7hqad/Nx1fv5m08/duzYER8fH7Kzs41t1dXV5OXlGdfQgoODqaqqwmq1GjVWq5WjR4861OTl5VFdXW3UZGdn065dOzp27AjAHXfc4fA6dTWnX6sTEZEr13mFWlVVFQUFBRQUFFBbW8uBAwcoKChg//79uLm5MW7cONLS0sjMzKSwsJCEhARatGhBbGwsAAEBAfTv35/ExES2bNmC1WolMTGRAQMGGGkeGxuLh4cHCQkJFBYWkpmZSVpaGgkJCcbqbuzYsXz55ZfMnTuX3bt3M3fuXHJychg3btwlmh4REWlMzuuW/pycHO6777562x988EHS09Ox2+2kpKTwxhtvYLPZ6NmzJ3PmzCEoKMioPXToEMnJyXz66acAREVFMXv2bDw9PY2aHTt2kJSUxNatW/H09CQuLo7k5GSHa2Zr1qxh5syZ7N27lz/84Q88++yz/OlPf/pNk3A50ukY19C8u4bmvWG6pf/CXfDvqcnvQ3/JXUPz7hqa94Yp1C6cnv0oIiKmoVATERHTUKiJiIhpKNRERMQ0FGoiImIaCjURETENhZqIiJiGQk1ERExDoSYiIqahUBMREdNQqImIiGko1ERExDQUaiIiYhoKNRERMQ2FmoiImIZCTURETEOhJiIipqFQExER01CoiYiIaSjURETENBRqIiJiGgo1ERExDaeEWk1NDTNnzuSWW27Bx8eHW265hZkzZ3Ly5Emjxm63M2vWLLp06YKvry/R0dHs3LnToR+bzUZ8fDx+fn74+fkRHx+PzWZzqNmxYweDBg3C19eXwMBAUlNTsdvtzjgMERFp5JwSamlpaSxbtozU1FSsVispKSksXbqUuXPnGjXz589n0aJFpKamkpWVhcViISYmhsrKSqNm9OjRFBQUsGrVKjIyMigoKGDMmDFG+5EjR4iJicHb25usrCxSUlJYsGABCxcudMZhiIhII9fUGZ1YrVYGDhxIVFQUAB07diQqKor/+Z//AU6t0tLT05k4cSKDBw8GID09HX9/fzIyMoiLi2PXrl1s2LCBdevWERISAsC8efOIioqiqKgIf39/Vq1axbFjx0hPT8fDw4OgoCB2797N4sWLGT9+PG5ubs44HBERaaScslILDQ0lNzeX3bt3A/Dtt9+Sk5PD3XffDUBxcTGlpaVEREQY+3h4eBAWFkZ+fj5wKhhbtmxpBFpdvy1atHCo6dWrFx4eHkZNZGQkJSUlFBcXO+NQRESkEXPKSm3ixIlUVVUREhKCu7s7J0+eJCkpidGjRwNQWloKgMVicdjPYrFQUlICQFlZGV5eXg6rLTc3N9q2bUtZWZlR0759+3p91LV16tSpwfEVFRX99oN0gcY67sZO8+4amveG3O7qAVy0S/l++vv7n7XNKaG2evVq3nvvPZYtW0aXLl345ptvmDJlCn5+fowcOdKoO/P0oN1urxdiZ/q1mrqbRM516vFcE3C5qjvlKr8vzbtraN7Nx1Xvp1NC7bnnnmP8+PHcf//9AHTt2pX9+/czb948Ro4ciY+PD3BqNdWhQwdjv/LycmOl5e3tTXl5uUOI2e12KioqHGrqVm2n9wH1V4EiInLlcco1tZ9//hl3d3eHbe7u7tTW1gKnbhzx8fEhOzvbaK+uriYvL8+4hhYcHExVVRVWq9WosVqtHD161KEmLy+P6upqoyY7O5t27drRsWNHZxyKiIg0Yk4JtYEDB5KWlsZnn31GcXExa9euZdGiRdx7773AqVOD48aNIy0tjczMTAoLC0lISKBFixbExsYCEBAQQP/+/UlMTGTLli1YrVYSExMZMGCAsYyNjY3Fw8ODhIQECgsLyczMJC0tjYSEBN35KCIizjn9OHv2bP72t7/x1FNPUV5ejo+PD6NGjWLy5MlGzYQJEzh27BiTJk3CZrPRs2dPVq9ezbXXXmvULF26lOTkZIYOHQpAVFQUs2fPNtpbt27Nhx9+SFJSEv369cPT05PHH3+c8ePHO+MwRESkkXOz2Wx6HMdlSBfOXUPz7hqa94Z5erZ29RAums122CWvq2c/ioiIaSjURETENBRqIiJiGgo1ERExDYWaiIiYhkJNRERMQ6EmIiKmoVATERHTUKiJiIhpKNRERMQ0FGoiImIaCjURETENhZqIiJiGQk1ERExDoSYiIqahUBMREdNQqImIiGko1ERExDQUaiIiYhoKNRERMQ2FmoiImIZCTURETMNpofbDDz8wduxYbrrpJnx8fAgJCSE3N9dot9vtzJo1iy5duuDr60t0dDQ7d+506MNmsxEfH4+fnx9+fn7Ex8djs9kcanbs2MGgQYPw9fUlMDCQ1NRU7Ha7sw5DREQaMaeEms1mY8CAAdjtdt5//33y8/OZPXs2FovFqJk/fz6LFi0iNTWVrKwsLBYLMTExVFZWGjWjR4+moKCAVatWkZGRQUFBAWPGjDHajxw5QkxMDN7e3mRlZZGSksKCBQtYuHChMw5DREQauabO6OTVV1/F19eXJUuWGNs6depk/Nlut5Oens7EiRMZPHgwAOnp6fj7+5ORkUFcXBy7du1iw4YNrFu3jpCQEADmzZtHVFQURUVF+Pv7s2rVKo4dO0Z6ejoeHh4EBQWxe/duFi9ezPjx43Fzc3PG4YiISCPllJXaxx9/TM+ePYmLi6Nz586Eh4fz97//3TgtWFxcTGlpKREREcZoAuTQAAAP90lEQVQ+Hh4ehIWFkZ+fD4DVaqVly5ZGoAGEhobSokULh5pevXrh4eFh1ERGRlJSUkJxcbEzDkVERBoxp6zU9u7dy/Lly0lISGDixIl88803JCcnAxAfH09paSmAw+nIuq9LSkoAKCsrw8vLy2G15ebmRtu2bSkrKzNq2rdvX6+PurbTV4enKyoq+u0H6QKNddyNnebdNTTvDbnd1QO4aJfy/fT39z9rm1NCrba2lh49ejB9+nQAbr31Vr777juWLVtGfHy8UXfm6UG73V4vxM70azV1q8FznXo81wRcrupOucrvS/PuGpp383HV++mU048+Pj4EBAQ4bLv55ps5cOCA0Q4YK6465eXlxkrL29ub8vJyhzsZ7XY7FRUVDjUN9QH1V4EiInLlcUqohYaGsmfPHodte/bs4YYbbgCgY8eO+Pj4kJ2dbbRXV1eTl5dnXEMLDg6mqqoKq9Vq1FitVo4ePepQk5eXR3V1tVGTnZ1Nu3bt6NixozMORUREGjGnhFpCQgJbtmxhzpw5fPfdd3z00Uf8/e9/Z/To0cCpU4Pjxo0jLS2NzMxMCgsLSUhIoEWLFsTGxgIQEBBA//79SUxMZMuWLVitVhITExkwYICxjI2NjcXDw4OEhAQKCwvJzMwkLS2NhIQE3fkoIiLOuaZ22223sXLlSl588UVefvllOnTowDPPPGOEGsCECRM4duwYkyZNwmaz0bNnT1avXs21115r1CxdupTk5GSGDh0KQFRUFLNnzzbaW7duzYcffkhSUhL9+vXD09OTxx9/nPHjxzvjMEREpJFzs9lsehzHZUgXzl1D8+4amveGeXq2dvUQLprNdtglr6tnP4qIiGko1ERExDQUaiIiYhoKNRERMQ2FmoiImIZCTURETEOhJiIipqFQExER03DKE0VEfi+X/pdRL91Hfbjql1FFriRaqYmIiGko1ERExDQUaiIiYhoKNRERMQ2FmoiImIZCTURETEOhJiIipqFQExER01CoiYiIaSjURETENBRqIiJiGgo1ERExDYWaiIiYxiUJtVdeeQVPT08mTZpkbLPb7cyaNYsuXbrg6+tLdHQ0O3fudNjPZrMRHx+Pn58ffn5+xMfHY7PZHGp27NjBoEGD8PX1JTAwkNTUVOx2+6U4DBERaWScHmpbtmzhzTffpGvXrg7b58+fz6JFi0hNTSUrKwuLxUJMTAyVlZVGzejRoykoKGDVqlVkZGRQUFDAmDFjjPYjR44QExODt7c3WVlZpKSksGDBAhYuXOjswxARkUbIqaF2+PBhHnvsMRYsWICnp6ex3W63k56ezsSJExk8eDBBQUGkp6dTVVVFRkYGALt27WLDhg2kpaUREhJCcHAw8+bN47PPPqOoqAiAVatWcezYMdLT0wkKCmLw4MFMmDCBxYsXa7UmIiLODbW60Orbt6/D9uLiYkpLS4mIiDC2eXh4EBYWRn5+PgBWq5WWLVsSEhJi1ISGhtKiRQuHml69euHh4WHUREZGUlJSQnFxsTMPRUREGiGnffL1m2++yXfffceSJUvqtZWWlgJgsVgctlssFkpKSgAoKyvDy8sLNzc3o93NzY22bdtSVlZm1LRv375eH3VtnTp1anBsdSu9xqaxjvvSunSfTH2p6f08N81PQ/T93hB/f/+ztjkl1IqKinjxxRf59NNPueqqq85ad3pgwanTkmeG2Jl+rabutGND+9Y51wRcroqKihrluOXs9H6enb7fzcdV76dTTj9arVYqKiro1asXXl5eeHl5sWnTJpYtW4aXlxfXXXcdgLHiqlNeXm6stLy9vSkvL3e4Nma326moqHCoaagPqL8KFBGRK49TQi06OprNmzeTk5Nj/NejRw/uv/9+cnJy6Ny5Mz4+PmRnZxv7VFdXk5eXZ1xDCw4OpqqqCqvVatRYrVaOHj3qUJOXl0d1dbVRk52dTbt27ejYsaMzDkVERBoxp5x+9PT0dLjbEaB58+a0adOGoKAgAMaNG8crr7yCv78/nTt3Zs6cObRo0YLY2FgAAgIC6N+/P4mJicyfPx+73U5iYiIDBgwwlrGxsbGkpqaSkJBAUlISe/bsIS0tjcmTJ5/z9KOIiFwZnHajyK+ZMGECx44dY9KkSdhsNnr27Mnq1au59tprjZqlS5eSnJzM0KFDAYiKimL27NlGe+vWrfnwww9JSkqiX79+eHp68vjjjzN+/Pjf6zBEROQy5maz2fQLXpchXThvmKdna1cP4aLZbIddPYTLlr7fG6bv9wunZz+KiIhpKNRERMQ0FGoiImIaCjURETENhZqIiJiGQk1ERExDoSYiIqbxu/3ytYg0Xpf+96Uu3dPo9fuBVxat1ERExDQUaiIiYhoKNRERMQ2FmoiImIZCTURETEOhJiIipqFQExER01CoiYiIaSjURETENBRqIiJiGgo1ERExDYWaiIiYhkJNRERMwymhNnfuXPr168cNN9zATTfdxPDhwyksLHSosdvtzJo1iy5duuDr60t0dDQ7d+50qLHZbMTHx+Pn54efnx/x8fHYbDaHmh07djBo0CB8fX0JDAwkNTUVu93ujMMQEZFGzimhlpuby6OPPspnn31GZmYmTZs2ZciQIRw6dMiomT9/PosWLSI1NZWsrCwsFgsxMTFUVlYaNaNHj6agoIBVq1aRkZFBQUEBY8aMMdqPHDlCTEwM3t7eZGVlkZKSwoIFC1i4cKEzDkNERBo5N5vN5vRlTlVVFX5+fqxcuZKoqCjsdjtdunThscceIykpCYBjx47h7+/PjBkziIuLY9euXYSEhLBu3TpCQ0MByMvLIyoqii1btuDv78/y5ct5/vnn2b17Nx4eHgC8/PLLrFixgsLCQtzc3Jx9KC5TVFSEv7+/q4dx2bn0n+t16TTmz/XSvLuG5v3CXZJralVVVdTW1uLp6QlAcXExpaWlREREGDUeHh6EhYWRn58PgNVqpWXLloSEhBg1oaGhtGjRwqGmV69eRqABREZGUlJSQnFx8aU4FBERaUQuySdfT5kyhW7duhEcHAxAaWkpABaLxaHOYrFQUlICQFlZGV5eXg6rLTc3N9q2bUtZWZlR0759+3p91LV16tTpUhxOg/RJwCIilx+nh9ozzzzDV199xbp163B3d3doO/P0oN1urxdiZ/q1mrqbRM516rGoqOj8D+C8XbrQudQuzXz8XjTvrqF5dw3Ne0POdWnGqaH29NNPs3r1atauXeuwavLx8QFOraY6dOhgbC8vLzdWWt7e3pSXlzuEmN1up6KiwqGmbtV2eh9QfxV4Ol2bcqT5cA3Nu2to3l3DVfPutGtqycnJZGRkkJmZyc033+zQ1rFjR3x8fMjOzja2VVdXk5eXZ1xDCw4OpqqqCqvVatRYrVaOHj3qUJOXl0d1dbVRk52dTbt27ejYsaOzDkVERBopp4RaUlIS77zzDsuWLcPT05PS0lJKS0upqqoCTp0aHDduHGlpaWRmZlJYWEhCQgItWrQgNjYWgICAAPr3709iYiJbtmzBarWSmJjIgAEDjMSPjY3Fw8ODhIQECgsLyczMJC0tjYSEBFPd+SgiIhfHKbf0193leKbk5GSefvpp4NSpxJSUFN544w1sNhs9e/Zkzpw5BAUFGfWHDh0iOTmZTz/9FICoqChmz57t0P+OHTtISkpi69ateHp6EhcXR3Jy8u8earrV1jU0766heXcNzfuFuyS/p3Yl0Deba2jeXUPz7hqa9wunZz+KiIhpKNRERMQ0FGoiImIaCjURETENhZqIiJiGQk1ERExDoSYiIqahUBMREdNQqImIiGko1ERExDQUaiIiYhoKNRERMQ2FmoiImIZCTURETEOhJiIipqFQExER01CoiYiIaSjURETENBRqIiJiGgo1ERExDYWaiIiYhkJNRERMo1GG2rJly7jlllvw8fGhb9++bN682dVDEhGRy0CjC7XVq1czZcoUnnrqKb788kuCg4N54IEH2L9/v6uHJiIiLtboQm3RokU89NBDjBo1ioCAAF5++WV8fHxYsWKFq4cmIiIu1tTVA7gQv/zyC9u3b+eJJ55w2B4REUF+fv7vOhab7fDv+npyiubdNTTvrqF5v3CNaqVWUVFBTU0NFovFYbvFYqGsrMxFoxIRkctFowq1Om5ubg5f2+32ettEROTK06hCzcvLC3d393qrsvLy8nqrNxERufI0qlC76qqr6N69O9nZ2Q7bs7OzCQkJcdGoRETkctGobhQBePzxxxkzZgw9e/YkJCSEFStW8MMPPxAXF+fqoYmIiIs1ulAbOnQoP/30Ey+//DKlpaUEBgby/vvv4+fn5+qhiYg0GidOnKBZs2auHobTudlsNrurByEiIr8vi8VCbm4uAQEBrh6KUzW6ldqV5sCBA8yaNYtFixa5eiimY7PZyM/Px9PTk+DgYIc7aI8ePcrChQtJTk524QjNqbCwkC1bthAcHExgYCDffvstixcv5vjx4wwfPpyIiAhXD9FUJk+e3OD2mpoaXnnlFTw9PQGYPXv27zmsS0YrtcvcN998Q9++ffnpp59cPRRT2blzJ0OGDKG8vJza2lpuvfVW3nrrLeM0dllZGV26dNG8O9nnn3/Oww8/TMuWLfn555/5xz/+wdixY+nWrRu1tbVs2rSJDz74gLvuusvVQzWNNm3a8Mc//pHWrVs7bN+0aRM9evSgefPmuLm5sXbtWheN0LkUai727rvvnrO9bqWmf1yda8SIETRt2pQlS5ZQWVnJlClTsFqtrF27lptuukmhdoncc8899OnTh2effZYPPviAp556ikcffZRp06YB8MILL7B9+3Y+/PBDF4/UPF555RXeeustFi5cyJ133mlsb9u2Lbm5uXTp0sWFo3M+hZqLtWnTxvhJqSG1tbVUV1frH1cn69y5M2vXriUwMNDY9swzz/Dhhx+ydu1aWrVqpVC7BPz8/Ni4cSM33ngjtbW1eHt7s2HDBrp37w6cOjU5ZMgQdu/e7eKRmsuWLVuIj48nJiaGqVOn4u7ubtpQa1S/p2ZG7dq1Iz09nQMHDjT437p161w9RFP65Zdf6v0g8dJLLzFkyBCio6PZtWuXi0Zmfk2aNDH+f8011xjXdABatmzJkSNHXDU007rjjjvYuHEj33//Pf3792fPnj2uHtIlo1BzsVtvvZWCgoKztru5uWG3azHtbJ07d2bbtm31ts+aNYshQ4bw8MMPu2BU5nfDDTfwn//8x/j6888/p0OHDsbXBw8exNvb2xVDM73WrVvz+uuv8+c//5mBAwdSW1vr6iFdErr70cWeeOIJjh49etb2G2+80TQXcC8n9957Lx988AEPPvhgvbbU1FRqampYvny5C0ZmbnFxcfzyyy/G10FBQQ7tn3/+OeHh4b/3sK4oo0aNonfv3litVtq3b+/q4TidrqmJiIhp6PSjiIiYhkJNRERMQ6EmIiKmoVATERHTUKiJiIhp/B9EyHwX8YifvAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "submission['Target'].value_counts().sort_index().plot.bar(color = 'blue');\n",
    "plt.title('Distribution of Predicted Labels for Individuals', size = 10);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data[data['Target'].notnull()]['Target'].value_counts().sort_index().plot.bar(color = 'blue');\n",
    "plt.title('Distribution of Labels for Training Individuals', size = 10);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Note:\n",
    "- The distribution of the training individuals and the targetted are very similar"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature Tools: Custom Primitive\n",
    "\n",
    "To expand the capabilities of featuretools, we can write our own primitives to be applied to the data. We'll write a simple function that finds the range of a numeric column.\n",
    "\n",
    "We construcing the dataframe, we include the aggregate and transmitive primitives."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Range\n",
    "from featuretools.primitives import make_agg_primitive\n",
    "\n",
    "# Custom primitive\n",
    "def range_calc(numeric):\n",
    "    return np.max(numeric) - np.min(numeric)\n",
    "\n",
    "range_ = make_agg_primitive(function = range_calc,\n",
    "                            input_types = [ft.variable_types.Numeric], \n",
    "                            return_type = ft.variable_types.Numeric)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Correlation Primitive\n",
    "def p_corr_calc(numeric1, numeric2):\n",
    "    return np.corrcoef(numeric1, numeric2)[0, 1]\n",
    "\n",
    "pcorr_ = make_agg_primitive(function = p_corr_calc,\n",
    "                            input_types = [ft.variable_types.Numeric, ft.variable_types.Numeric], \n",
    "                            return_type = ft.variable_types.Numeric)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes:\n",
    "- `entityset`: The entity we created above with the relationship among the table.\n",
    "- `target_entity`: Entity id of entity on which to make predictions -  the household.\n",
    "- `agg_primitives`: Aggregation types that will be implemented.\n",
    "- `trans_primitives`: Transformative types that will be implemented.\n",
    "- `drop_exact`: Drop features that exactly match these strings in name.\n",
    "    - We are dropping the columns from this dataset. We will later add to our current dataset.\n",
    "- `max_depth`: Maximum allowed depth of features. It creates columns from columns the columns it creates.\n",
    "- `verbose`: Should it print out the process.\n",
    "    - Verbosity in keyword arguments usually means showing more 'wordy' information for the task.\n",
    "- `n_jobs`: number of parallel processes to use when calculating feature matrix. \n",
    "    - When the argument is set to -1, it uses all the cores at 100%.\n",
    "- `chunk_size`: Number of rows of output feature matrix to calculate at time. If passed an integer greater than 0, will try to use that many rows per chunk. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "feature_matrix_add, feature_names_add = ft.dfs(entityset=es, target_entity = 'households', \n",
    "                                               agg_primitives=['min', 'max', 'mean', 'percent_true', 'all', 'any',\n",
    "                                                               'sum', 'skew', 'std', range_],\n",
    "                                               trans_primitives=[], drop_exact=all_features,\n",
    "                                               max_depth=2, verbose=1, n_jobs=-1, chunk_size=100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes:\n",
    "- `%%capture`\n",
    "    - This simply discards the output, but the %%capture magic can be used to save the output to a variable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<Feature: Target>,\n",
       " <Feature: MIN(data.tech)>,\n",
       " <Feature: MEAN(data.tech)>,\n",
       " <Feature: ALL(data.v18q)>,\n",
       " <Feature: ALL(data.dis)>]"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# First 5 of the Feature names to be added in our dataframe\n",
    "feature_names_add[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "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>Target</th>\n",
       "      <th>MIN(data.tech)</th>\n",
       "      <th>MEAN(data.tech)</th>\n",
       "      <th>ALL(data.v18q)</th>\n",
       "      <th>ALL(data.dis)</th>\n",
       "      <th>ALL(data.male)</th>\n",
       "      <th>ALL(data.estadocivil1)</th>\n",
       "      <th>ALL(data.estadocivil2)</th>\n",
       "      <th>ALL(data.estadocivil3)</th>\n",
       "      <th>ALL(data.estadocivil4)</th>\n",
       "      <th>...</th>\n",
       "      <th>ANY(data.instlevel8)</th>\n",
       "      <th>ANY(data.instlevel9)</th>\n",
       "      <th>ANY(data.mobilephone)</th>\n",
       "      <th>ANY(data.rez_esc-missing)</th>\n",
       "      <th>RANGE_CALC(data.escolari)</th>\n",
       "      <th>RANGE_CALC(data.rez_esc)</th>\n",
       "      <th>RANGE_CALC(data.age)</th>\n",
       "      <th>RANGE_CALC(data.escolari/age)</th>\n",
       "      <th>RANGE_CALC(data.instl_var/age)</th>\n",
       "      <th>RANGE_CALC(data.tech)</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>idhogar</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>000a08204</th>\n",
       "      <td>NaN</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "      <td>26</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>000bce7c4</th>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>0.083333</td>\n",
       "      <td>0.016667</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001845fb0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>8</td>\n",
       "      <td>0.0</td>\n",
       "      <td>33</td>\n",
       "      <td>0.410931</td>\n",
       "      <td>0.128205</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001ff74ca</th>\n",
       "      <td>4.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>16</td>\n",
       "      <td>0.0</td>\n",
       "      <td>38</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>003123ec2</th>\n",
       "      <td>2.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>23</td>\n",
       "      <td>0.304348</td>\n",
       "      <td>0.130435</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 75 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           Target  MIN(data.tech)  MEAN(data.tech)  ALL(data.v18q)  \\\n",
       "idhogar                                                              \n",
       "000a08204     NaN               2                2            True   \n",
       "000bce7c4     NaN               1                1           False   \n",
       "001845fb0     NaN               1                1           False   \n",
       "001ff74ca     4.0               2                2            True   \n",
       "003123ec2     2.0               1                1           False   \n",
       "\n",
       "           ALL(data.dis)  ALL(data.male)  ALL(data.estadocivil1)  \\\n",
       "idhogar                                                            \n",
       "000a08204          False           False                   False   \n",
       "000bce7c4          False           False                   False   \n",
       "001845fb0          False           False                   False   \n",
       "001ff74ca          False           False                   False   \n",
       "003123ec2          False           False                   False   \n",
       "\n",
       "           ALL(data.estadocivil2)  ALL(data.estadocivil3)  \\\n",
       "idhogar                                                     \n",
       "000a08204                   False                   False   \n",
       "000bce7c4                   False                    True   \n",
       "001845fb0                   False                   False   \n",
       "001ff74ca                   False                   False   \n",
       "003123ec2                   False                   False   \n",
       "\n",
       "           ALL(data.estadocivil4)          ...            \\\n",
       "idhogar                                    ...             \n",
       "000a08204                   False          ...             \n",
       "000bce7c4                   False          ...             \n",
       "001845fb0                   False          ...             \n",
       "001ff74ca                   False          ...             \n",
       "003123ec2                   False          ...             \n",
       "\n",
       "           ANY(data.instlevel8)  ANY(data.instlevel9)  ANY(data.mobilephone)  \\\n",
       "idhogar                                                                        \n",
       "000a08204                  True                 False                   True   \n",
       "000bce7c4                 False                 False                   True   \n",
       "001845fb0                  True                 False                   True   \n",
       "001ff74ca                  True                 False                   True   \n",
       "003123ec2                 False                 False                   True   \n",
       "\n",
       "           ANY(data.rez_esc-missing)  RANGE_CALC(data.escolari)  \\\n",
       "idhogar                                                           \n",
       "000a08204                      False                         14   \n",
       "000bce7c4                      False                          5   \n",
       "001845fb0                       True                          8   \n",
       "001ff74ca                      False                         16   \n",
       "003123ec2                      False                          7   \n",
       "\n",
       "           RANGE_CALC(data.rez_esc)  RANGE_CALC(data.age)  \\\n",
       "idhogar                                                     \n",
       "000a08204                       0.0                    26   \n",
       "000bce7c4                       0.0                     3   \n",
       "001845fb0                       0.0                    33   \n",
       "001ff74ca                       0.0                    38   \n",
       "003123ec2                       0.0                    23   \n",
       "\n",
       "           RANGE_CALC(data.escolari/age)  RANGE_CALC(data.instl_var/age)  \\\n",
       "idhogar                                                                    \n",
       "000a08204                       0.500000                        0.250000   \n",
       "000bce7c4                       0.083333                        0.016667   \n",
       "001845fb0                       0.410931                        0.128205   \n",
       "001ff74ca                       0.000000                        0.000000   \n",
       "003123ec2                       0.304348                        0.130435   \n",
       "\n",
       "           RANGE_CALC(data.tech)  \n",
       "idhogar                           \n",
       "000a08204                      0  \n",
       "000bce7c4                      0  \n",
       "001845fb0                      0  \n",
       "001ff74ca                      0  \n",
       "003123ec2                      0  \n",
       "\n",
       "[5 rows x 75 columns]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Dataframe that will be added to our data\n",
    "feature_matrix_add.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10340, 245)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Adding the features that were created from the feature tools process to our all features list\n",
    "all_features += [str(x.get_name()) for x in feature_names_add]\n",
    "\n",
    "# Adding the df that were created from the feature tools process to our feature_matrix\n",
    "feature_matrix = pd.concat([feature_matrix, feature_matrix_add], axis = 1)\n",
    "feature_matrix.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Post Processing Function\n",
    "\n",
    "There are a number of steps after generating the feature matrix so let's put all of these in a function.\n",
    "- Remove any duplicated columns.\n",
    "- Replace infinite values with np.nan\n",
    "- Remove columns with a missing percentage above the missing_threshold\n",
    "- Remove columns with only a single unique value.\n",
    "- Remove one out of every pair of columns with a correlation threshold above the correlation_threshold\n",
    "- Extract the training and testing data along with labels and ids (needed for making submissions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "def post_process(feature_matrix,\n",
    "                 missing_threshold = 0.95, \n",
    "                 correlation_threshold = 0.95):\n",
    "    \n",
    "    # Remove duplicated features\n",
    "    start_features = feature_matrix.shape[1]\n",
    "    feature_matrix = feature_matrix.iloc[:, ~feature_matrix.columns.duplicated()]\n",
    "    n_duplicated = start_features - feature_matrix.shape[1]\n",
    "    print(f'There were {n_duplicated} duplicated features.')\n",
    "    \n",
    "    feature_matrix = feature_matrix.replace({np.inf: np.nan, -np.inf:np.nan}).reset_index()\n",
    "    \n",
    "    # Remove the ids and labels\n",
    "    ids = list(feature_matrix.pop('idhogar'))\n",
    "    labels = list(feature_matrix.pop('Target'))\n",
    "    \n",
    "    # Remove columns derived from the Target\n",
    "    drop_cols = []\n",
    "    for col in feature_matrix:\n",
    "        if col == 'Target':\n",
    "            pass\n",
    "        else:\n",
    "            if 'Target' in col:\n",
    "                drop_cols.append(col)\n",
    "                \n",
    "    feature_matrix = feature_matrix[[x for x in feature_matrix if x not in drop_cols]] \n",
    "    \n",
    "    # One hot encoding (if necessary)\n",
    "    feature_matrix = pd.get_dummies(feature_matrix)\n",
    "    n_features_start = feature_matrix.shape[1]\n",
    "    print('Original shape: ', feature_matrix.shape)\n",
    "    \n",
    "    # Find missing and percentage\n",
    "    missing = pd.DataFrame(feature_matrix.isnull().sum())\n",
    "    missing['fraction'] = missing[0] / feature_matrix.shape[0]\n",
    "    missing.sort_values('fraction', ascending = False, inplace = True)\n",
    "\n",
    "    # Missing above threshold\n",
    "    missing_cols = list(missing[missing['fraction'] > missing_threshold].index)\n",
    "    n_missing_cols = len(missing_cols)\n",
    "\n",
    "    # Remove missing columns\n",
    "    feature_matrix = feature_matrix[[x for x in feature_matrix if x not in missing_cols]]\n",
    "    print('{} missing columns with threshold: {}.'.format(n_missing_cols, missing_threshold))\n",
    "    \n",
    "    # Zero variance\n",
    "    unique_counts = pd.DataFrame(feature_matrix.nunique()).sort_values(0, ascending = True)\n",
    "    zero_variance_cols = list(unique_counts[unique_counts[0] == 1].index)\n",
    "    n_zero_variance_cols = len(zero_variance_cols)\n",
    "\n",
    "    # Remove zero variance columns\n",
    "    feature_matrix = feature_matrix[[x for x in feature_matrix if x not in zero_variance_cols]]\n",
    "    print('{} zero variance columns.'.format(n_zero_variance_cols))\n",
    "    \n",
    "    # Correlations\n",
    "    corr_matrix = feature_matrix.corr()\n",
    "\n",
    "    # Extract the upper triangle of the correlation matrix\n",
    "    upper = corr_matrix.where(np.triu(np.ones(corr_matrix.shape), k = 1).astype(np.bool))\n",
    "\n",
    "    # Select the features with correlations above the threshold\n",
    "    # Need to use the absolute value\n",
    "    to_drop = [column for column in upper.columns if any(upper[column].abs() > correlation_threshold)]\n",
    "\n",
    "    n_collinear = len(to_drop)\n",
    "    \n",
    "    feature_matrix = feature_matrix[[x for x in feature_matrix if x not in to_drop]]\n",
    "    print('{} collinear columns removed with correlation above {}.'.format(n_collinear,  correlation_threshold))\n",
    "    \n",
    "    total_removed = n_duplicated, n_missing_cols + n_zero_variance_cols + n_collinear\n",
    "    \n",
    "    print('Total columns removed: ', total_removed)\n",
    "    print('Shape after feature selection: {}.'.format(feature_matrix.shape))\n",
    "    \n",
    "    # Extract the ids and labels\n",
    "    feature_matrix['idhogar'] = ids\n",
    "    feature_matrix['Target'] = labels\n",
    "    \n",
    "    # Extract out training and testing data\n",
    "    train = feature_matrix[feature_matrix['Target'].notnull()]\n",
    "    test = feature_matrix[feature_matrix['Target'].isnull()]\n",
    "    \n",
    "    # Subset to houses with a head of household\n",
    "    train = train[train['idhogar'].isin(list(train_valid['idhogar']))]\n",
    "    test = test[test['idhogar'].isin(list(test_valid['idhogar']))]\n",
    "    \n",
    "    # Training labels and testing household ids\n",
    "    train_labels = np.array(train.pop('Target')).reshape((-1,))\n",
    "    test_ids = list(test.pop('idhogar'))\n",
    "    \n",
    "    # Align the dataframes to ensure they have the same columns\n",
    "    train, test = train.align(test, join = 'inner', axis = 1)\n",
    "    \n",
    "    assert(len(train_labels) == train.shape[0]), \"Labels must be same length as number of training observations\"\n",
    "    assert(len(test_ids) == test.shape[0]), \"Must be equal number of test ids as testing observations\"\n",
    "    \n",
    "    return train, train_labels, test, test_ids"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There were 1 duplicated features.\n",
      "Original shape:  (10340, 243)\n",
      "0 missing columns with threshold: 0.95.\n",
      "11 zero variance columns.\n",
      "14 collinear columns removed with correlation above 0.95.\n",
      "Total columns removed:  (1, 25)\n",
      "Shape after feature selection: (10340, 218).\n"
     ]
    }
   ],
   "source": [
    "train, train_labels, test, test_ids = post_process(feature_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model: Post Feature Processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Fold 1, Validation Score: 0.42849, Estimators Trained: 180'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 2, Validation Score: 0.42569, Estimators Trained: 22'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 3, Validation Score: 0.40739, Estimators Trained: 11'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 4, Validation Score: 0.40639, Estimators Trained: 16'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 5, Validation Score: 0.39958, Estimators Trained: 7'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'5 cross validation score: 0.41351 with std: 0.01144.'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%capture --no-display\n",
    "\n",
    "submission, feature_importances, valid_scores = model_gbm(train, train_labels, test, test_ids, 5)\n",
    "results = results.append(pd.DataFrame({'version': ['additional_5fold'], \n",
    "                                       'F1-mean': [valid_scores.mean()], \n",
    "                                       'F1-std': [valid_scores.std()]}))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes:\n",
    "- Performance prior to the new features was \"5 cross validation score: 0.42273 with std: 0.02513.\" Hence, performance worsen when we included more features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xa287228d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_feature_importances(feature_importances)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  Increase `max_features`\n",
    "- `drop_contains`: Drop features that contains these strings in name."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "feature_matrix_add, feature_names_add = ft.dfs(entityset=es, target_entity = 'households', \n",
    "                                               agg_primitives=['min', 'max', 'mean', 'percent_true', 'all', 'any',\n",
    "                                                               'sum', 'skew', 'std', range_, pcorr_],\n",
    "                                               trans_primitives=[], drop_contains=list(all_features),\n",
    "                                               max_depth= 2, max_features=1000, verbose=1, \n",
    "                                               n_jobs=-1, chunk_size=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There were 1 duplicated features.\n",
      "Original shape:  (10340, 273)\n",
      "10 missing columns with threshold: 0.95.\n",
      "11 zero variance columns.\n",
      "24 collinear columns removed with correlation above 0.95.\n",
      "Total columns removed:  (1, 45)\n",
      "Shape after feature selection: (10340, 228).\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(2973, 228)"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Add additional names to our all_features list\n",
    "all_features += [str(x.get_name()) for x in feature_names_add]\n",
    "\n",
    "# Concat the feature tools dataframe with our feature_matrix datraframe\n",
    "feature_matrix = pd.concat([feature_matrix, feature_matrix_add], axis = 1)\n",
    "\n",
    "# Updating our dataset to remove any duplicates, null columns, no variance, or collinearity columns\n",
    "train, train_labels, test, test_ids = post_process(feature_matrix)\n",
    "train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original training shape (2973, 228)\n",
      "Training shape after removing zero importance features (2973, 141)\n"
     ]
    }
   ],
   "source": [
    "print('Original training shape', train.shape)\n",
    "train = train[list(feature_importances.loc[feature_importances['importance'] != 0, 'feature'])]\n",
    "test = test[train.columns]\n",
    "\n",
    "print('Training shape after removing zero importance features', train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Add in `Divide` Primitive\n",
    "- Next we'll add a divide transform primitive into the deep feature synthesis call. At first we'll limit the features to 1000."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "feature_matrix_add, feature_names_add = ft.dfs(entityset=es, target_entity = 'households', \n",
    "                                       agg_primitives = ['min', 'max', 'mean', 'percent_true', 'all', 'any',\n",
    "                                                         'sum', 'skew', 'std', range_, pcorr_],\n",
    "                                       trans_primitives = ['divide_numeric'], drop_contains = list(all_features),\n",
    "                                       max_depth = 2, max_features = 1000,\n",
    "                                       verbose = 1, n_jobs = -1, \n",
    "                                       chunk_size = 1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There were 1 duplicated features.\n",
      "Original shape:  (10340, 1273)\n",
      "437 missing columns with threshold: 0.95.\n",
      "11 zero variance columns.\n",
      "446 collinear columns removed with correlation above 0.95.\n",
      "Total columns removed:  (1, 894)\n",
      "Shape after feature selection: (10340, 379).\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(2973, 379)"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_features += [str(x.get_name()) for x in feature_names_add]\n",
    "feature_matrix = pd.concat([feature_matrix, feature_matrix_add], axis = 1)\n",
    "train, train_labels, test, test_ids = post_process(feature_matrix)\n",
    "train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Fold 1, Validation Score: 0.43425, Estimators Trained: 99'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 2, Validation Score: 0.43282, Estimators Trained: 24'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 3, Validation Score: 0.38534, Estimators Trained: 23'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 4, Validation Score: 0.42623, Estimators Trained: 4'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 5, Validation Score: 0.40903, Estimators Trained: 15'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'5 cross validation score: 0.41754 with std: 0.01842.'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%capture --no-display\n",
    "\n",
    "submission, feature_importances, valid_scores = model_gbm(train, train_labels, test, test_ids, 5)\n",
    "results = results.append(pd.DataFrame({'version': ['divide1000_5fold'], \n",
    "                                       'F1-mean': [valid_scores.mean()], \n",
    "                                       'F1-std': [valid_scores.std()]}))\n",
    "#submission.to_csv('divide1000_featuretools.csv', index = False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes:\n",
    "- Performance prior to the new features was \"5 cross validation score: 0.42273 with std: 0.02513.\" Hence, performance worsen (0.41754) when we included more features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_feature_importances(feature_importances)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Modeling\n",
    "\n",
    "We will be working with several models which will allow us to see which model offers better performance. We can also time the performance... to understand the trade-off between performance and computation efficiency.\n",
    "\n",
    "Overview of models:\n",
    "1. Linear SVC: A Linear Suppport Vector Machine. Without digging into the details, a support vector machine tries to split the data values by its target variable optimizing towards maximizing the margin between the target variables. If we were to have two potential target variables, you can think of drawing a line between the data values. Mathematically, the line is calculated by having the greatest margin between the data values of each potential target variable and the line known as the Decision Boundary.\n",
    "2. GaussianNB: Gaussian Naive Bayes is build from the Bayes Theorem. The model is trained where it uses probability. For example, if you are predicting a person's level of poverty, we would look at the historical probability. If the new person has a car and historically data indicates that 12% of people with cars have \n",
    "3. MLPClassifier: A Multilayer perceptron is an artificial neural network. Meaning, it consists of at least three layers (input, hidden, and output). A neural network is like creating many functions. Within each function, through backward propogation, the model will optimized the weights or betas.\n",
    "4. LogisticRegressionCV: Logistic Regression is a probalisitic method that assigns a probability to each potential target variable. A logistic regression is a unique function because it \"squishes\" the value to be between 0 and 1.\n",
    "5. RidgeClassifierCV: TBD\n",
    "6. LinearDiscriminantAnalysis: The Linear Discriminant Analysis is a two-part computation. It reduces the features used while maximizing the difference between the mean of the target variable. However, it must also attempt to minimize the variance of each target variable.\n",
    "7. KNeigbhorsClassifier: K-Neigbors is a model that focuses on using the features that are geometrically close to the data value you are predicting. If calculated on a 2-D graph, we will compute the Euclidean distance to get the x number of neigbors. It then uses the target variable of the data values that are close to the prediction data value.\n",
    "8. ExtraTreeClassifier: Similar to a Random Forest, it uses Decision Trees to buid it model. Unlike a Random Forest, an Extra Tree Classifier does not relay on bootstrapping the dataset. Rather, it uses the entire dataset. One advantage of using the entire dataset is that it reduces bias because with the model is build with more data. More data can help a model reduce bias by introducing more features!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Notes: \n",
    "- Based on the performance of three dataset produced by `featuretools`, I learned that the first dataset (code below) drives the best performance.\n",
    "- We will be using the first dataset produced by `featuretools` to create fine-tune the performance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Built 173 features\n",
      "EntitySet scattered to workers in 5.797 seconds\n",
      "\n",
      "\n",
      "\n",
      "Elapsed: 00:00 | Remaining: ? | Progress:   0%|          | Calculated: 0/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:03 | Remaining: 06:21 | Progress:   1%|          | Calculated: 1/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:03 | Remaining: 04:20 | Progress:   4%|▍         | Calculated: 4/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:06 | Remaining: 04:32 | Progress:   5%|▍         | Calculated: 5/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:07 | Remaining: 03:08 | Progress:   7%|▋         | Calculated: 7/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:07 | Remaining: 02:15 | Progress:   8%|▊         | Calculated: 8/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:11 | Remaining: 03:23 | Progress:   9%|▊         | Calculated: 9/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:11 | Remaining: 02:22 | Progress:  11%|█         | Calculated: 11/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:11 | Remaining: 01:43 | Progress:  12%|█▏        | Calculated: 12/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:15 | Remaining: 03:08 | Progress:  12%|█▎        | Calculated: 13/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:20 | Remaining: 02:34 | Progress:  16%|█▋        | Calculated: 17/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:20 | Remaining: 01:49 | Progress:  18%|█▊        | Calculated: 19/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:20 | Remaining: 01:21 | Progress:  19%|█▉        | Calculated: 20/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:23 | Remaining: 01:57 | Progress:  20%|██        | Calculated: 21/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:23 | Remaining: 01:23 | Progress:  22%|██▏       | Calculated: 23/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:23 | Remaining: 01:05 | Progress:  23%|██▎       | Calculated: 24/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:26 | Remaining: 01:40 | Progress:  24%|██▍       | Calculated: 25/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:26 | Remaining: 01:12 | Progress:  26%|██▌       | Calculated: 27/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:26 | Remaining: 00:57 | Progress:  27%|██▋       | Calculated: 28/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:29 | Remaining: 01:33 | Progress:  28%|██▊       | Calculated: 29/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:29 | Remaining: 01:07 | Progress:  30%|██▉       | Calculated: 31/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:29 | Remaining: 00:54 | Progress:  31%|███       | Calculated: 32/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:32 | Remaining: 01:27 | Progress:  32%|███▏      | Calculated: 33/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:32 | Remaining: 01:04 | Progress:  34%|███▎      | Calculated: 35/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:32 | Remaining: 00:49 | Progress:  35%|███▍      | Calculated: 36/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:35 | Remaining: 01:20 | Progress:  36%|███▌      | Calculated: 37/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:35 | Remaining: 00:58 | Progress:  38%|███▊      | Calculated: 39/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:35 | Remaining: 00:45 | Progress:  38%|███▊      | Calculated: 40/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:38 | Remaining: 01:13 | Progress:  39%|███▉      | Calculated: 41/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:38 | Remaining: 00:54 | Progress:  40%|████      | Calculated: 42/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:38 | Remaining: 00:43 | Progress:  41%|████▏     | Calculated: 43/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:38 | Remaining: 00:35 | Progress:  42%|████▏     | Calculated: 44/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:41 | Remaining: 01:02 | Progress:  43%|████▎     | Calculated: 45/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:41 | Remaining: 00:45 | Progress:  44%|████▍     | Calculated: 46/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:41 | Remaining: 00:39 | Progress:  45%|████▌     | Calculated: 47/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:41 | Remaining: 00:32 | Progress:  46%|████▌     | Calculated: 48/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:44 | Remaining: 01:01 | Progress:  47%|████▋     | Calculated: 49/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:44 | Remaining: 00:45 | Progress:  49%|████▉     | Calculated: 51/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:45 | Remaining: 00:34 | Progress:  50%|█████     | Calculated: 52/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:47 | Remaining: 00:58 | Progress:  51%|█████     | Calculated: 53/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:47 | Remaining: 00:42 | Progress:  53%|█████▎    | Calculated: 55/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:47 | Remaining: 00:32 | Progress:  54%|█████▍    | Calculated: 56/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:50 | Remaining: 00:51 | Progress:  55%|█████▍    | Calculated: 57/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:50 | Remaining: 00:38 | Progress:  57%|█████▋    | Calculated: 59/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:50 | Remaining: 00:30 | Progress:  58%|█████▊    | Calculated: 60/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:52 | Remaining: 00:46 | Progress:  59%|█████▊    | Calculated: 61/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:53 | Remaining: 00:34 | Progress:  61%|██████    | Calculated: 63/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:53 | Remaining: 00:26 | Progress:  62%|██████▏   | Calculated: 64/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:55 | Remaining: 00:42 | Progress:  62%|██████▎   | Calculated: 65/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:56 | Remaining: 00:31 | Progress:  64%|██████▍   | Calculated: 67/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:56 | Remaining: 00:24 | Progress:  65%|██████▌   | Calculated: 68/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:58 | Remaining: 00:38 | Progress:  66%|██████▋   | Calculated: 69/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:59 | Remaining: 00:28 | Progress:  68%|██████▊   | Calculated: 71/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 00:59 | Remaining: 00:21 | Progress:  69%|██████▉   | Calculated: 72/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:01 | Remaining: 00:36 | Progress:  70%|███████   | Calculated: 73/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:02 | Remaining: 00:26 | Progress:  72%|███████▏  | Calculated: 75/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:02 | Remaining: 00:21 | Progress:  73%|███████▎  | Calculated: 76/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:05 | Remaining: 00:30 | Progress:  74%|███████▍  | Calculated: 77/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:05 | Remaining: 00:21 | Progress:  75%|███████▌  | Calculated: 78/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:06 | Remaining: 00:22 | Progress:  76%|███████▌  | Calculated: 79/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:07 | Remaining: 00:21 | Progress:  77%|███████▋  | Calculated: 80/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:10 | Remaining: 00:39 | Progress:  78%|███████▊  | Calculated: 81/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:11 | Remaining: 00:28 | Progress:  80%|███████▉  | Calculated: 83/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:12 | Remaining: 00:23 | Progress:  81%|████████  | Calculated: 84/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:15 | Remaining: 00:32 | Progress:  82%|████████▏ | Calculated: 85/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:16 | Remaining: 00:22 | Progress:  84%|████████▎ | Calculated: 87/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:17 | Remaining: 00:19 | Progress:  85%|████████▍ | Calculated: 88/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:20 | Remaining: 00:26 | Progress:  86%|████████▌ | Calculated: 89/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:20 | Remaining: 00:18 | Progress:  87%|████████▋ | Calculated: 90/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:21 | Remaining: 00:15 | Progress:  88%|████████▊ | Calculated: 91/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:22 | Remaining: 00:13 | Progress:  88%|████████▊ | Calculated: 92/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:25 | Remaining: 00:19 | Progress:  89%|████████▉ | Calculated: 93/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:25 | Remaining: 00:13 | Progress:  90%|█████████ | Calculated: 94/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:26 | Remaining: 00:10 | Progress:  91%|█████████▏| Calculated: 95/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:27 | Remaining: 00:08 | Progress:  92%|█████████▏| Calculated: 96/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:29 | Remaining: 00:08 | Progress:  93%|█████████▎| Calculated: 97/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:29 | Remaining: 00:05 | Progress:  94%|█████████▍| Calculated: 98/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:30 | Remaining: 00:04 | Progress:  95%|█████████▌| Calculated: 99/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:30 | Remaining: 00:03 | Progress:  96%|█████████▌| Calculated: 100/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:32 | Remaining: 00:03 | Progress:  97%|█████████▋| Calculated: 101/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:32 | Remaining: 00:00 | Progress:  99%|█████████▉| Calculated: 103/104 chunks\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "Elapsed: 01:33 | Remaining: 00:00 | Progress: 100%|██████████| Calculated: 104/104 chunks\u001b[A\u001b[A\u001b[A"
     ]
    }
   ],
   "source": [
    "# Deep Feature Synthesis: Aggregation and Transform primitives are build as default\n",
    "# The max_depth parameter controls the maximum depth of the features returned by DFS\n",
    "# With a depth of 2, a number of features are generated using the supplied primitives\n",
    "feature_matrix, feature_names = ft.dfs(entityset=es, \n",
    "                                       target_entity='households', \n",
    "                                       max_depth=2, \n",
    "                                       verbose=1, \n",
    "                                       n_jobs=-1, \n",
    "                                       chunk_size=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "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>hacdor</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>paredblolad</th>\n",
       "      <th>paredzocalo</th>\n",
       "      <th>paredpreb</th>\n",
       "      <th>pisocemento</th>\n",
       "      <th>pareddes</th>\n",
       "      <th>paredmad</th>\n",
       "      <th>...</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel2)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel3)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel4)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel5)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel6)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel7)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel8)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel9)</th>\n",
       "      <th>PERCENT_TRUE(data.mobilephone)</th>\n",
       "      <th>PERCENT_TRUE(data.rez_esc-missing)</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>idhogar</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>000a08204</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>000bce7c4</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001845fb0</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001ff74ca</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>003123ec2</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 173 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           hacdor  hacapo  v14a  refrig  paredblolad  paredzocalo  paredpreb  \\\n",
       "idhogar                                                                        \n",
       "000a08204   False   False  True    True         True        False      False   \n",
       "000bce7c4   False   False  True    True        False         True      False   \n",
       "001845fb0   False   False  True    True         True        False      False   \n",
       "001ff74ca   False   False  True    True         True        False      False   \n",
       "003123ec2   False   False  True    True        False        False       True   \n",
       "\n",
       "           pisocemento  pareddes  paredmad  \\\n",
       "idhogar                                      \n",
       "000a08204        False     False     False   \n",
       "000bce7c4        False     False     False   \n",
       "001845fb0        False     False     False   \n",
       "001ff74ca        False     False     False   \n",
       "003123ec2         True     False     False   \n",
       "\n",
       "                          ...                  PERCENT_TRUE(data.instlevel2)  \\\n",
       "idhogar                   ...                                                  \n",
       "000a08204                 ...                                            0.0   \n",
       "000bce7c4                 ...                                            0.5   \n",
       "001845fb0                 ...                                            0.0   \n",
       "001ff74ca                 ...                                            0.0   \n",
       "003123ec2                 ...                                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel3)  PERCENT_TRUE(data.instlevel4)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                           0.00   \n",
       "000bce7c4                           0.00                           0.00   \n",
       "001845fb0                           0.25                           0.25   \n",
       "001ff74ca                           0.00                           0.00   \n",
       "003123ec2                           0.25                           0.25   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel5)  PERCENT_TRUE(data.instlevel6)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                            0.0   \n",
       "000bce7c4                           0.00                            0.0   \n",
       "001845fb0                           0.25                            0.0   \n",
       "001ff74ca                           0.00                            0.0   \n",
       "003123ec2                           0.00                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel7)  PERCENT_TRUE(data.instlevel8)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                            0.0                       0.666667   \n",
       "000bce7c4                            0.0                       0.000000   \n",
       "001845fb0                            0.0                       0.250000   \n",
       "001ff74ca                            0.0                       0.500000   \n",
       "003123ec2                            0.0                       0.000000   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel9)  PERCENT_TRUE(data.mobilephone)  \\\n",
       "idhogar                                                                    \n",
       "000a08204                            0.0                             1.0   \n",
       "000bce7c4                            0.0                             1.0   \n",
       "001845fb0                            0.0                             1.0   \n",
       "001ff74ca                            0.0                             1.0   \n",
       "003123ec2                            0.0                             1.0   \n",
       "\n",
       "           PERCENT_TRUE(data.rez_esc-missing)  \n",
       "idhogar                                        \n",
       "000a08204                                0.00  \n",
       "000bce7c4                                0.00  \n",
       "001845fb0                                0.25  \n",
       "001ff74ca                                0.00  \n",
       "003123ec2                                0.00  \n",
       "\n",
       "[5 rows x 173 columns]"
      ]
     },
     "execution_count": 104,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_features = [str(x.get_name()) for x in feature_names]\n",
    "feature_matrix.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>hacdor</th>\n",
       "      <th>hacapo</th>\n",
       "      <th>v14a</th>\n",
       "      <th>refrig</th>\n",
       "      <th>paredblolad</th>\n",
       "      <th>paredzocalo</th>\n",
       "      <th>paredpreb</th>\n",
       "      <th>pisocemento</th>\n",
       "      <th>pareddes</th>\n",
       "      <th>paredmad</th>\n",
       "      <th>...</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel2)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel3)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel4)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel5)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel6)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel7)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel8)</th>\n",
       "      <th>PERCENT_TRUE(data.instlevel9)</th>\n",
       "      <th>PERCENT_TRUE(data.mobilephone)</th>\n",
       "      <th>PERCENT_TRUE(data.rez_esc-missing)</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>idhogar</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>000a08204</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>000bce7c4</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001845fb0</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>001ff74ca</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>003123ec2</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.25</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 173 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           hacdor  hacapo  v14a  refrig  paredblolad  paredzocalo  paredpreb  \\\n",
       "idhogar                                                                        \n",
       "000a08204   False   False  True    True         True        False      False   \n",
       "000bce7c4   False   False  True    True        False         True      False   \n",
       "001845fb0   False   False  True    True         True        False      False   \n",
       "001ff74ca   False   False  True    True         True        False      False   \n",
       "003123ec2   False   False  True    True        False        False       True   \n",
       "\n",
       "           pisocemento  pareddes  paredmad  \\\n",
       "idhogar                                      \n",
       "000a08204        False     False     False   \n",
       "000bce7c4        False     False     False   \n",
       "001845fb0        False     False     False   \n",
       "001ff74ca        False     False     False   \n",
       "003123ec2         True     False     False   \n",
       "\n",
       "                          ...                  PERCENT_TRUE(data.instlevel2)  \\\n",
       "idhogar                   ...                                                  \n",
       "000a08204                 ...                                            0.0   \n",
       "000bce7c4                 ...                                            0.5   \n",
       "001845fb0                 ...                                            0.0   \n",
       "001ff74ca                 ...                                            0.0   \n",
       "003123ec2                 ...                                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel3)  PERCENT_TRUE(data.instlevel4)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                           0.00   \n",
       "000bce7c4                           0.00                           0.00   \n",
       "001845fb0                           0.25                           0.25   \n",
       "001ff74ca                           0.00                           0.00   \n",
       "003123ec2                           0.25                           0.25   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel5)  PERCENT_TRUE(data.instlevel6)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                           0.00                            0.0   \n",
       "000bce7c4                           0.00                            0.0   \n",
       "001845fb0                           0.25                            0.0   \n",
       "001ff74ca                           0.00                            0.0   \n",
       "003123ec2                           0.00                            0.0   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel7)  PERCENT_TRUE(data.instlevel8)  \\\n",
       "idhogar                                                                   \n",
       "000a08204                            0.0                       0.666667   \n",
       "000bce7c4                            0.0                       0.000000   \n",
       "001845fb0                            0.0                       0.250000   \n",
       "001ff74ca                            0.0                       0.500000   \n",
       "003123ec2                            0.0                       0.000000   \n",
       "\n",
       "           PERCENT_TRUE(data.instlevel9)  PERCENT_TRUE(data.mobilephone)  \\\n",
       "idhogar                                                                    \n",
       "000a08204                            0.0                             1.0   \n",
       "000bce7c4                            0.0                             1.0   \n",
       "001845fb0                            0.0                             1.0   \n",
       "001ff74ca                            0.0                             1.0   \n",
       "003123ec2                            0.0                             1.0   \n",
       "\n",
       "           PERCENT_TRUE(data.rez_esc-missing)  \n",
       "idhogar                                        \n",
       "000a08204                                0.00  \n",
       "000bce7c4                                0.00  \n",
       "001845fb0                                0.25  \n",
       "001ff74ca                                0.00  \n",
       "003123ec2                                0.00  \n",
       "\n",
       "[5 rows x 173 columns]"
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Remove any columns that are a derivations of the target variable \n",
    "# These are created because some of transform primitives might have affected the Target.\n",
    "\n",
    "drop_cols = []\n",
    "for col in feature_matrix:\n",
    "    if col == 'Target':\n",
    "        pass\n",
    "    else:\n",
    "        if 'Target' in col:\n",
    "            drop_cols.append(col)\n",
    "            \n",
    "print(drop_cols)            \n",
    "feature_matrix = feature_matrix[[x for x in feature_matrix if x not in drop_cols]]         \n",
    "feature_matrix.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Remove features that have an absolute inter-correlation value is greater than 0.99 \n",
    "\n",
    "corr_matrix, to_drop = get_corrs_matrix(feature_matrix, corr_threshold=.99)\n",
    "feature_matrix = feature_matrix[[x for x in feature_matrix if x not in to_drop]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = feature_matrix[feature_matrix['Target'].notnull()].reset_index()\n",
    "test = feature_matrix[feature_matrix['Target'].isnull()].reset_index()\n",
    "\n",
    "# Removing rows from training set we do not have idhogar\n",
    "train = train[train['idhogar'].isin(list(train_valid['idhogar']))]\n",
    "train.head()\n",
    "\n",
    "# Removing rows from testing set where we do not have idhogar\n",
    "test = test[test['idhogar'].isin(list(test_valid['idhogar']))]\n",
    "test.head()\n",
    "\n",
    "# Labels for training\n",
    "train_labels = np.array(train.pop('Target')).reshape(-1, )\n",
    "test_ids = list(test.pop('idhogar'))\n",
    "\n",
    "# Aligning the columns\n",
    "train, test = train.align(test, axis = 1, join = 'inner')\n",
    "all_features = list(train.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Fold 1, Validation Score: 0.4165, Estimators Trained: 23'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 2, Validation Score: 0.45101, Estimators Trained: 75'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 3, Validation Score: 0.42141, Estimators Trained: 9'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 4, Validation Score: 0.38444, Estimators Trained: 36'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 5, Validation Score: 0.4402, Estimators Trained: 45'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'5 cross validation score: 0.42271 with std: 0.02285.'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%capture --no-display\n",
    "submission, feature_importances, valid_scores = model_gbm(train, train_labels, test, test_ids, 5)\n",
    "\n",
    "results = pd.DataFrame({'version': ['default_5fold'], \n",
    "                        'F1-mean': [valid_scores.mean()], \n",
    "                        'F1-std': [valid_scores.std()]})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Fold 1, Validation Score: 0.41231, Estimators Trained: 7'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 2, Validation Score: 0.45219, Estimators Trained: 24'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 3, Validation Score: 0.47161, Estimators Trained: 17'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 4, Validation Score: 0.42306, Estimators Trained: 53'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 5, Validation Score: 0.48986, Estimators Trained: 44'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 6, Validation Score: 0.42653, Estimators Trained: 32'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 7, Validation Score: 0.41745, Estimators Trained: 51'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 8, Validation Score: 0.44556, Estimators Trained: 212'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 9, Validation Score: 0.39433, Estimators Trained: 22'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Fold 10, Validation Score: 0.41544, Estimators Trained: 17'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'10 cross validation score: 0.43483 with std: 0.02798.'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.08s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.12s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.12s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.12s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.10s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.15s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.17s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.18s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.10s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.21s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.22s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.22s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.12s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.27s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.27s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.28s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.10s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.28s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.28s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.29s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.10s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.11s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.10s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.31s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.31s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.31s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.31s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.32s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.32s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.31s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.32s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 5292.33s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.58s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.59s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.64s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.65s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.72s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.61s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.61s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.63s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.62s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.59s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.60s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.60s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.89s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.88s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.89s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.88s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.90s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.89s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.89s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.90s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.89s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.90s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.90s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.93s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.93s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.93s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.92s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.93s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.92s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.92s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "tornado.application - ERROR - Exception in callback <bound method SystemMonitor.update of <SystemMonitor: cpu: 17 memory: 134 MB fds: 141>>\n",
      "Traceback (most recent call last):\n",
      "  File \"/Users/alexguanga/anaconda3/lib/python3.6/site-packages/tornado/ioloop.py\", line 1229, in _run\n",
      "    return self.callback()\n",
      "  File \"/Users/alexguanga/anaconda3/lib/python3.6/site-packages/distributed/system_monitor.py\", line 72, in update\n",
      "    read_bytes = (ioc.bytes_recv - last.bytes_recv) / (duration or 0.5)\n",
      "AttributeError: 'NoneType' object has no attribute 'bytes_recv'\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.97s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n",
      "distributed.core - INFO - Event loop was unresponsive in Nanny for 28284.97s.  This is often caused by long-running GIL-holding functions or moving large chunks of data. This can cause timeouts and instability.\n"
     ]
    }
   ],
   "source": [
    "%%capture --no-display\n",
    "submission, feature_importances, valid_scores = model_gbm(train, train_labels, test, test_ids, 10)\n",
    "\n",
    "results = pd.DataFrame({'version': ['default_10fold'], \n",
    "                        'F1-mean': [valid_scores.mean()], \n",
    "                        'F1-std': [valid_scores.std()]})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture --no-display\n",
    "submission, feature_importances, valid_scores = model_gbm(train, train_labels, test, test_ids, 20)\n",
    "\n",
    "results = pd.DataFrame({'version': ['default_10fold'], \n",
    "                        'F1-mean': [valid_scores.mean()], \n",
    "                        'F1-std': [valid_scores.std()]})"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}