3.1_Prepr_train1_imputation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data preprocessing: Imputation of missing values\n",
    "\n",
    "**NOTEBOOK GOAL**: Impute missing values\n",
    "\n",
    "**DATASET TRANSFORMATION**: `preprocessed0_date_(train|test).csv` >> `preprocessed1_imputed_(train|test).csv`\n",
    "\n",
    "**INDEX**\n",
    "- [Missing values](#Missing-values)\n",
    "- [Max_Gust_SpeedKm_h attribute processing](#Max_Gust_SpeedKm_h-attribute-processing)\n",
    "- [Events attribute processing](#Events-attribute-processing)\n",
    "- [NumberOfCustomers, NumberOfSales attributes processing](#NumberOfCustomers,-NumberOfSales-attributes-processing)\n",
    "- [NumberOfCustomers attribute processing](#NumberOfCustomers-attribute-processing)\n",
    "- [CloudCover attribute processing](#CloudCover-attribute-processing)\n",
    "- [Min_VisibilitykM, Mean_VisibilityKm and-Max_VisibilityKm attributes processing](#Min_VisibilitykM,-Mean_VisibilityKm-and-Max_VisibilityKm-attributes-processing)\n",
    "- [Final check](#Final-check)\n",
    "- [Write to file](#Write-to-file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "# reoder variables to work on train or test dataset\n",
    "work_on = 'test'\n",
    "#work_on = 'train'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "from import_man import *\n",
    "\n",
    "sns.set(style=\"white\", color_codes=True)\n",
    "sns.set_context(rc={\"font.family\":'sans',\"font.size\":24,\"axes.titlesize\":24,\"axes.labelsize\":24})  \n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "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>StoreID</th>\n",
       "      <th>Date</th>\n",
       "      <th>IsHoliday</th>\n",
       "      <th>IsOpen</th>\n",
       "      <th>HasPromotions</th>\n",
       "      <th>StoreType</th>\n",
       "      <th>AssortmentType</th>\n",
       "      <th>NearestCompetitor</th>\n",
       "      <th>Region</th>\n",
       "      <th>Region_AreaKM2</th>\n",
       "      <th>...</th>\n",
       "      <th>Min_Sea_Level_PressurehPa</th>\n",
       "      <th>Min_TemperatureC</th>\n",
       "      <th>Min_VisibilitykM</th>\n",
       "      <th>Precipitationmm</th>\n",
       "      <th>WindDirDegrees</th>\n",
       "      <th>D_Day</th>\n",
       "      <th>D_DayOfYear</th>\n",
       "      <th>D_Month</th>\n",
       "      <th>D_Year</th>\n",
       "      <th>D_DayOfweek</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1000</td>\n",
       "      <td>01/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1011</td>\n",
       "      <td>2</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>180</td>\n",
       "      <td>1</td>\n",
       "      <td>60</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1000</td>\n",
       "      <td>02/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1009</td>\n",
       "      <td>3</td>\n",
       "      <td>10.0</td>\n",
       "      <td>5.08</td>\n",
       "      <td>315</td>\n",
       "      <td>2</td>\n",
       "      <td>61</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1000</td>\n",
       "      <td>03/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1013</td>\n",
       "      <td>-2</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>210</td>\n",
       "      <td>3</td>\n",
       "      <td>62</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1000</td>\n",
       "      <td>04/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1002</td>\n",
       "      <td>1</td>\n",
       "      <td>6.0</td>\n",
       "      <td>3.05</td>\n",
       "      <td>193</td>\n",
       "      <td>4</td>\n",
       "      <td>63</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1000</td>\n",
       "      <td>05/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1000</td>\n",
       "      <td>2</td>\n",
       "      <td>8.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>247</td>\n",
       "      <td>5</td>\n",
       "      <td>64</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 39 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   StoreID        Date  IsHoliday  IsOpen  HasPromotions     StoreType  \\\n",
       "0     1000  01/03/2018          0       1              0  Hyper Market   \n",
       "1     1000  02/03/2018          0       1              0  Hyper Market   \n",
       "2     1000  03/03/2018          0       1              0  Hyper Market   \n",
       "3     1000  04/03/2018          0       0              0  Hyper Market   \n",
       "4     1000  05/03/2018          0       1              1  Hyper Market   \n",
       "\n",
       "  AssortmentType  NearestCompetitor  Region  Region_AreaKM2     ...       \\\n",
       "0        General                326       7            9643     ...        \n",
       "1        General                326       7            9643     ...        \n",
       "2        General                326       7            9643     ...        \n",
       "3        General                326       7            9643     ...        \n",
       "4        General                326       7            9643     ...        \n",
       "\n",
       "   Min_Sea_Level_PressurehPa  Min_TemperatureC  Min_VisibilitykM  \\\n",
       "0                       1011                 2              10.0   \n",
       "1                       1009                 3              10.0   \n",
       "2                       1013                -2              10.0   \n",
       "3                       1002                 1               6.0   \n",
       "4                       1000                 2               8.0   \n",
       "\n",
       "  Precipitationmm  WindDirDegrees  D_Day  D_DayOfYear  D_Month  D_Year  \\\n",
       "0            0.00             180      1           60        3    2018   \n",
       "1            5.08             315      2           61        3    2018   \n",
       "2            0.00             210      3           62        3    2018   \n",
       "3            3.05             193      4           63        3    2018   \n",
       "4            0.25             247      5           64        3    2018   \n",
       "\n",
       "   D_DayOfweek  \n",
       "0            3  \n",
       "1            4  \n",
       "2            5  \n",
       "3            6  \n",
       "4            0  \n",
       "\n",
       "[5 rows x 39 columns]"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# start form preprocessed0_date_train dataset\n",
    "df = pd.read_csv('./dataset/preprocessed0_date_' + work_on + '.csv')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Missing values\n",
    "\n",
    "Let's check how many missing values are in the data set and how can we deal with them. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "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 Ratio</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Max_Gust_SpeedKm_h</th>\n",
       "      <td>75.374817</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Events</th>\n",
       "      <td>35.984591</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CloudCover</th>\n",
       "      <td>18.424566</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Min_VisibilitykM</th>\n",
       "      <td>5.736611</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Mean_VisibilityKm</th>\n",
       "      <td>5.736611</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Max_VisibilityKm</th>\n",
       "      <td>5.736611</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    Missing Ratio\n",
       "Max_Gust_SpeedKm_h      75.374817\n",
       "Events                  35.984591\n",
       "CloudCover              18.424566\n",
       "Min_VisibilitykM         5.736611\n",
       "Mean_VisibilityKm        5.736611\n",
       "Max_VisibilityKm         5.736611"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_na = (df.isnull().sum() / len(df)) * 100\n",
    "df_na = df_na.drop(df_na[df_na == 0].index).sort_values(ascending=False)[:30]\n",
    "missing_data = pd.DataFrame({'Missing Ratio': df_na})\n",
    "missing_data.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Max_Gust_SpeedKm_h attribute processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The attribute Max_Gust_SpeedKm_h has too many missing values... We have 2 possibility:\n",
    "\n",
    "1. supposing that the missing values mean no Gust and replace them with 0\n",
    "2. drop the column and do not take it into account for training\n",
    "\n",
    "\n",
    "Let's consider the option n. 1 for the moment:\n",
    "\n",
    "Let's see in the training set if there is any correlation between the number of Sales wrt 'Max_Gust_SpeedKm_h' when we impute the missing value with '0'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_first = df\n",
    "df_first['Max_Gust_SpeedKm_h'] = df_first['Max_Gust_SpeedKm_h'].fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "if work_on == 'train':\n",
    "    plt.rcParams['figure.figsize'] = [10.0,6.0]\n",
    "    plt.rcParams['figure.dpi']=80\n",
    "    covariance = df_first[['NumberOfSales','NumberOfCustomers','Max_Gust_SpeedKm_h']].corr(method='pearson')\n",
    "    sns.heatmap(covariance, square=True, annot=True, cmap=\"Blues\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "'Max_Gust_SpeedKm_h seems to have no correlation with the targets.\n",
    "Considering the fact that the 80% of the values have been imputed following the first option, the best option seems to be not consider that attribute for training.\n",
    "Let's drop it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Columns before (39): Index(['StoreID', 'Date', 'IsHoliday', 'IsOpen', 'HasPromotions', 'StoreType',\n",
      "       'AssortmentType', 'NearestCompetitor', 'Region', 'Region_AreaKM2',\n",
      "       'Region_GDP', 'Region_PopulationK', 'CloudCover', 'Events',\n",
      "       'Max_Dew_PointC', 'Max_Gust_SpeedKm_h', 'Max_Humidity',\n",
      "       'Max_Sea_Level_PressurehPa', 'Max_TemperatureC', 'Max_VisibilityKm',\n",
      "       'Max_Wind_SpeedKm_h', 'Mean_Dew_PointC', 'Mean_Humidity',\n",
      "       'Mean_Sea_Level_PressurehPa', 'Mean_TemperatureC', 'Mean_VisibilityKm',\n",
      "       'Mean_Wind_SpeedKm_h', 'Min_Dew_PointC', 'Min_Humidity',\n",
      "       'Min_Sea_Level_PressurehPa', 'Min_TemperatureC', 'Min_VisibilitykM',\n",
      "       'Precipitationmm', 'WindDirDegrees', 'D_Day', 'D_DayOfYear', 'D_Month',\n",
      "       'D_Year', 'D_DayOfweek'],\n",
      "      dtype='object')\n",
      "Columns before (38): Index(['StoreID', 'Date', 'IsHoliday', 'IsOpen', 'HasPromotions', 'StoreType',\n",
      "       'AssortmentType', 'NearestCompetitor', 'Region', 'Region_AreaKM2',\n",
      "       'Region_GDP', 'Region_PopulationK', 'CloudCover', 'Events',\n",
      "       'Max_Dew_PointC', 'Max_Humidity', 'Max_Sea_Level_PressurehPa',\n",
      "       'Max_TemperatureC', 'Max_VisibilityKm', 'Max_Wind_SpeedKm_h',\n",
      "       'Mean_Dew_PointC', 'Mean_Humidity', 'Mean_Sea_Level_PressurehPa',\n",
      "       'Mean_TemperatureC', 'Mean_VisibilityKm', 'Mean_Wind_SpeedKm_h',\n",
      "       'Min_Dew_PointC', 'Min_Humidity', 'Min_Sea_Level_PressurehPa',\n",
      "       'Min_TemperatureC', 'Min_VisibilitykM', 'Precipitationmm',\n",
      "       'WindDirDegrees', 'D_Day', 'D_DayOfYear', 'D_Month', 'D_Year',\n",
      "       'D_DayOfweek'],\n",
      "      dtype='object')\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>StoreID</th>\n",
       "      <th>Date</th>\n",
       "      <th>IsHoliday</th>\n",
       "      <th>IsOpen</th>\n",
       "      <th>HasPromotions</th>\n",
       "      <th>StoreType</th>\n",
       "      <th>AssortmentType</th>\n",
       "      <th>NearestCompetitor</th>\n",
       "      <th>Region</th>\n",
       "      <th>Region_AreaKM2</th>\n",
       "      <th>...</th>\n",
       "      <th>Min_Sea_Level_PressurehPa</th>\n",
       "      <th>Min_TemperatureC</th>\n",
       "      <th>Min_VisibilitykM</th>\n",
       "      <th>Precipitationmm</th>\n",
       "      <th>WindDirDegrees</th>\n",
       "      <th>D_Day</th>\n",
       "      <th>D_DayOfYear</th>\n",
       "      <th>D_Month</th>\n",
       "      <th>D_Year</th>\n",
       "      <th>D_DayOfweek</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1000</td>\n",
       "      <td>01/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1011</td>\n",
       "      <td>2</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>180</td>\n",
       "      <td>1</td>\n",
       "      <td>60</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1000</td>\n",
       "      <td>02/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1009</td>\n",
       "      <td>3</td>\n",
       "      <td>10.0</td>\n",
       "      <td>5.08</td>\n",
       "      <td>315</td>\n",
       "      <td>2</td>\n",
       "      <td>61</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1000</td>\n",
       "      <td>03/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1013</td>\n",
       "      <td>-2</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>210</td>\n",
       "      <td>3</td>\n",
       "      <td>62</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1000</td>\n",
       "      <td>04/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1002</td>\n",
       "      <td>1</td>\n",
       "      <td>6.0</td>\n",
       "      <td>3.05</td>\n",
       "      <td>193</td>\n",
       "      <td>4</td>\n",
       "      <td>63</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1000</td>\n",
       "      <td>05/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1000</td>\n",
       "      <td>2</td>\n",
       "      <td>8.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>247</td>\n",
       "      <td>5</td>\n",
       "      <td>64</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 38 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   StoreID        Date  IsHoliday  IsOpen  HasPromotions     StoreType  \\\n",
       "0     1000  01/03/2018          0       1              0  Hyper Market   \n",
       "1     1000  02/03/2018          0       1              0  Hyper Market   \n",
       "2     1000  03/03/2018          0       1              0  Hyper Market   \n",
       "3     1000  04/03/2018          0       0              0  Hyper Market   \n",
       "4     1000  05/03/2018          0       1              1  Hyper Market   \n",
       "\n",
       "  AssortmentType  NearestCompetitor  Region  Region_AreaKM2     ...       \\\n",
       "0        General                326       7            9643     ...        \n",
       "1        General                326       7            9643     ...        \n",
       "2        General                326       7            9643     ...        \n",
       "3        General                326       7            9643     ...        \n",
       "4        General                326       7            9643     ...        \n",
       "\n",
       "   Min_Sea_Level_PressurehPa  Min_TemperatureC  Min_VisibilitykM  \\\n",
       "0                       1011                 2              10.0   \n",
       "1                       1009                 3              10.0   \n",
       "2                       1013                -2              10.0   \n",
       "3                       1002                 1               6.0   \n",
       "4                       1000                 2               8.0   \n",
       "\n",
       "  Precipitationmm  WindDirDegrees  D_Day  D_DayOfYear  D_Month  D_Year  \\\n",
       "0            0.00             180      1           60        3    2018   \n",
       "1            5.08             315      2           61        3    2018   \n",
       "2            0.00             210      3           62        3    2018   \n",
       "3            3.05             193      4           63        3    2018   \n",
       "4            0.25             247      5           64        3    2018   \n",
       "\n",
       "   D_DayOfweek  \n",
       "0            3  \n",
       "1            4  \n",
       "2            5  \n",
       "3            6  \n",
       "4            0  \n",
       "\n",
       "[5 rows x 38 columns]"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# drop column Max_Gust_SpeedKm_h\n",
    "print(\"Columns before ({}): {}\".format(len(list(df.columns)), df.columns))\n",
    "\n",
    "df = df.drop(['Max_Gust_SpeedKm_h'], axis=1)\n",
    "\n",
    "print(\"Columns before ({}): {}\".format(len(list(df.columns)), df.columns))\n",
    "\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As a final outcome: **attribute Max_Gust_SpeedKm_h dropped**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Events attribute processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that Max_Gust_SpeedKm_h is gone let's continue with the imputation.\n",
    "\n",
    "The next attribute is 'Events' with a missing ratio of 24%.\n",
    "If we analyze the Events we notice that values like 'sunny' never appear so the best option seems to assume that missing values mean no particulare forcasting events.\n",
    "Let's fill them with the value 'none'."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "df[\"Events\"] = df[\"Events\"].fillna(\"None\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### NumberOfCustomers, NumberOfSales attributes processing\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's set the NumberOfCustomers and NumberOfSales to None when the store is closed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "if work_on == 'train':\n",
    "    df.loc[df['IsOpen'] == 0, ['NumberOfCustomers','NumberOfSales']] = 0\n",
    "    df.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Impute NumberOfSales\n",
    "#df[\"NumberOfSales\"] = df.groupby('StoreID')[\"NumberOfSales\"].transform(lambda x: x.fillna(x.mean()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Impute NumberOfCustomers\n",
    "#df[\"NumberOfCustomers\"] = df.groupby('StoreID')[\"NumberOfCustomers\"].transform(lambda x: x.fillna(x.mean()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.options.mode.chained_assignment = None  # default='warn'\n",
    "\n",
    "def convert_to_date(sample):\n",
    "    sample['Date'] = pd.to_datetime(sample['Date'],format='%d/%m/%Y')\n",
    "    \n",
    "def get_store_data(store_id):\n",
    "    return df.loc[df['StoreID'] == store_id]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.dates as mdates\n",
    "\n",
    "years = mdates.YearLocator()   # every year\n",
    "months = mdates.MonthLocator()  # every month\n",
    "monthsFmt = mdates.DateFormatter('%m')\n",
    "yearsFmt = mdates.DateFormatter('%Y')\n",
    "monthsYearFmt = mdates.DateFormatter('%m-%Y')\n",
    "\n",
    "#function that plots a store sales trend\n",
    "def plot_storeNo(store_id):\n",
    "    sampleStore = get_store_data(store_id)\n",
    "    convert_to_date(sampleStore)\n",
    "    \n",
    "    \n",
    "    \n",
    "    fig, ax = plt.subplots()\n",
    "    \n",
    "    # format the ticks\n",
    "    ax.xaxis.set_major_locator(months)\n",
    "    ax.xaxis.set_major_formatter(monthsYearFmt)\n",
    "    ax.xaxis.set_minor_locator(months)\n",
    "    \n",
    "    ax.plot(sampleStore.Date, sampleStore.NumberOfSales)\n",
    "    ax.xaxis.set_major_locator(months)\n",
    "    ax.xaxis.set_major_formatter(monthsYearFmt)\n",
    "    ax.xaxis.set_minor_locator(months)\n",
    "    ax.grid(True)\n",
    "    fig.autofmt_xdate()\n",
    "    plt.show()\n",
    "    \n",
    "#function that plots many store sales trends in the whole amount of time with legend\n",
    "def plot_stores(start_index,end_index):  \n",
    "    fig, ax = plt.subplots()\n",
    "\n",
    "    # format the ticks\n",
    "    ax.xaxis.set_major_locator(months)\n",
    "    ax.xaxis.set_major_formatter(monthsYearFmt)\n",
    "    ax.xaxis.set_minor_locator(months)\n",
    "    \n",
    "    stores_legend = []\n",
    "\n",
    "    for i in range(start_index,end_index):\n",
    "        sampleStore = get_store_data(i)\n",
    "        convert_to_date(sampleStore)\n",
    "    \n",
    "        ax.plot(sampleStore.Date, sampleStore.NumberOfSales)\n",
    "        stores_legend.append('Store '+str(i))\n",
    "        \n",
    "    ax.xaxis.set_major_locator(months)\n",
    "    ax.xaxis.set_major_formatter(monthsYearFmt)\n",
    "    ax.xaxis.set_minor_locator(months)\n",
    "    plt.legend(stores_legend, loc='upper left')\n",
    "\n",
    "    ax.grid(True)\n",
    "    fig.autofmt_xdate()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "sampleStore = df.loc[df['StoreID'] == 1000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "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>StoreID</th>\n",
       "      <th>Date</th>\n",
       "      <th>IsHoliday</th>\n",
       "      <th>IsOpen</th>\n",
       "      <th>HasPromotions</th>\n",
       "      <th>StoreType</th>\n",
       "      <th>AssortmentType</th>\n",
       "      <th>NearestCompetitor</th>\n",
       "      <th>Region</th>\n",
       "      <th>Region_AreaKM2</th>\n",
       "      <th>...</th>\n",
       "      <th>Min_Sea_Level_PressurehPa</th>\n",
       "      <th>Min_TemperatureC</th>\n",
       "      <th>Min_VisibilitykM</th>\n",
       "      <th>Precipitationmm</th>\n",
       "      <th>WindDirDegrees</th>\n",
       "      <th>D_Day</th>\n",
       "      <th>D_DayOfYear</th>\n",
       "      <th>D_Month</th>\n",
       "      <th>D_Year</th>\n",
       "      <th>D_DayOfweek</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1000</td>\n",
       "      <td>01/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1011</td>\n",
       "      <td>2</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>180</td>\n",
       "      <td>1</td>\n",
       "      <td>60</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1000</td>\n",
       "      <td>02/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1009</td>\n",
       "      <td>3</td>\n",
       "      <td>10.0</td>\n",
       "      <td>5.08</td>\n",
       "      <td>315</td>\n",
       "      <td>2</td>\n",
       "      <td>61</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1000</td>\n",
       "      <td>03/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1013</td>\n",
       "      <td>-2</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>210</td>\n",
       "      <td>3</td>\n",
       "      <td>62</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1000</td>\n",
       "      <td>04/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1002</td>\n",
       "      <td>1</td>\n",
       "      <td>6.0</td>\n",
       "      <td>3.05</td>\n",
       "      <td>193</td>\n",
       "      <td>4</td>\n",
       "      <td>63</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>4 rows × 38 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   StoreID        Date  IsHoliday  IsOpen  HasPromotions     StoreType  \\\n",
       "0     1000  01/03/2018          0       1              0  Hyper Market   \n",
       "1     1000  02/03/2018          0       1              0  Hyper Market   \n",
       "2     1000  03/03/2018          0       1              0  Hyper Market   \n",
       "3     1000  04/03/2018          0       0              0  Hyper Market   \n",
       "\n",
       "  AssortmentType  NearestCompetitor  Region  Region_AreaKM2     ...       \\\n",
       "0        General                326       7            9643     ...        \n",
       "1        General                326       7            9643     ...        \n",
       "2        General                326       7            9643     ...        \n",
       "3        General                326       7            9643     ...        \n",
       "\n",
       "   Min_Sea_Level_PressurehPa  Min_TemperatureC  Min_VisibilitykM  \\\n",
       "0                       1011                 2              10.0   \n",
       "1                       1009                 3              10.0   \n",
       "2                       1013                -2              10.0   \n",
       "3                       1002                 1               6.0   \n",
       "\n",
       "  Precipitationmm  WindDirDegrees  D_Day  D_DayOfYear  D_Month  D_Year  \\\n",
       "0            0.00             180      1           60        3    2018   \n",
       "1            5.08             315      2           61        3    2018   \n",
       "2            0.00             210      3           62        3    2018   \n",
       "3            3.05             193      4           63        3    2018   \n",
       "\n",
       "   D_DayOfweek  \n",
       "0            3  \n",
       "1            4  \n",
       "2            5  \n",
       "3            6  \n",
       "\n",
       "[4 rows x 38 columns]"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sampleStore[:4]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "if work_on == 'train':  \n",
    "    import datetime\n",
    "    import random\n",
    "    plt.rcParams['figure.figsize'] = [20.0,8.0]\n",
    "    plt.rcParams['figure.dpi']=80\n",
    "    \n",
    "    #x = sampleStore['Date'][:100]\n",
    "    #y = sampleStore['NumberOfSales'][:100]\n",
    "    x = sampleStore['Date']\n",
    "    y = sampleStore['NumberOfSales']\n",
    "    \n",
    "    plt.xlabel(\"Days\")\n",
    "    plt.ylabel(\"Sales\")\n",
    "    \n",
    "    colors = np.random.rand(*x.shape)\n",
    "    # plot\n",
    "    plt.scatter(x, y, c=colors, marker=r'$\\diamondsuit$', alpha=0.8)\n",
    "    # beautify the x-labels\n",
    "    plt.gcf().autofmt_xdate()\n",
    "    \n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "if work_on == 'train':\n",
    "    plot_storeNo(1023)\n",
    "    plot_storeNo(1001)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Imputation results checks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "if work_on == 'train':\n",
    "    matplotlib.rcParams['figure.figsize'] = (8.0, 6.0)\n",
    "# fit the data with a normal distribution\n",
    "    sns.distplot(df['NumberOfCustomers'] , fit=norm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x6832453be0>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x68203acb70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if work_on == 'train':    \n",
    "    matplotlib.rcParams['figure.figsize'] = (8.0, 6.0)\n",
    "    # fit the data with a normal distribution\n",
    "    sns.distplot(df['NumberOfCustomers'] , fit=norm)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we don't have anymore instances where the Store is closed. We visualize again the distribution of 'NumberOfCustomers'\n",
    "\n",
    "We visualize also the distribution of 'NumberOfSales' considering that the instancies where the Store was closed are already been removed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "if work_on == 'train':   \n",
    "    matplotlib.rcParams['figure.figsize'] = (8.0, 6.0)\n",
    "    # fit the data with a normal distribution\n",
    "    sns.distplot(df['NumberOfSales'] , fit=norm)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### CloudCover attribute processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Similarly, noticing that in CloudCover the values range from 1 to 9 but on the library is written that have a range 0-9 we assume that the missing values mean 0 level of covering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "df[\"CloudCover\"] = df[\"CloudCover\"].fillna(0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Min_VisibilitykM, Mean_VisibilityKm and Max_VisibilityKm attributes processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's analyse Min_VisibilitykM, Mean_VisibilityKm and Max_VisibilityKm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of Min_VisibilitykM.isnull\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "2621"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Min_Visibility_missing = df[df.Min_VisibilitykM.isnull()]\n",
    "print(\"number of Min_VisibilitykM.isnull\")\n",
    "Min_Visibility_missing.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min_VisibilitykM.isnull AND Max_VisibilityKm.isnull\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "2621"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# we compute the intersection between instances where Min_Visibilityand and Max_Visibility are missing\n",
    "Max_Visibility_missing = df[df.Max_VisibilityKm.isnull()]\n",
    "vis = Min_Visibility_missing.merge(Max_Visibility_missing)\n",
    "print(\"Min_VisibilitykM.isnull AND Max_VisibilityKm.isnull\")\n",
    "vis.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min_VisibilitykM.isnull AND Max_VisibilityKm.isnull AND Mean_VisibilityKm.isnull\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "2621"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# we intersect vis with the instances where Mean_Visibility is missing too\n",
    "Mean_Visibility_missing = df[df.Mean_VisibilityKm.isnull()]\n",
    "vis = vis.merge(Mean_Visibility_missing)\n",
    "\n",
    "print(\"Min_VisibilitykM.isnull AND Max_VisibilityKm.isnull AND Mean_VisibilityKm.isnull\")\n",
    "vis.shape[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "we notice that alla the dataframes have the same count. This means that when one of these 3 atributes is missing also the other ones are missing.\n",
    "We try to understand if the missing triples focus on a particoular region/period/Store because, if that, we have to deal with them carefully to not indroduce bias in the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "region 0, number of missing triples:\n",
      "55\n",
      "region 1, number of missing triples:\n",
      "48\n",
      "region 2, number of missing triples:\n",
      "1625\n",
      "region 3, number of missing triples:\n",
      "92\n",
      "region 4, number of missing triples:\n",
      "0\n",
      "region 5, number of missing triples:\n",
      "204\n",
      "region 6, number of missing triples:\n",
      "180\n"
     ]
    }
   ],
   "source": [
    "# plt.figure(figsize=(8, 6))\n",
    "# sns.set_context(\"poster\", font_scale=1.5, rc={\"lines.linewidth\": 2.5})\n",
    "a = iter(list(range(7)))\n",
    "for i in a:\n",
    "   # vis_gb=vis.groupby(['Region']).count()\n",
    "    vis_gb = vis[vis.Region == i]\n",
    "    print('region %d, number of missing triples:' % i )\n",
    "    print(vis_gb.shape[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "the missing triples do not focus on a single region, now we have to impute them.\n",
    "\n",
    "To do so, we use the **median of the region in that day**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "df[\"Mean_VisibilityKm\"] = df.groupby([\"Region\", \"Date\"])[\"Mean_VisibilityKm\"].transform(lambda x: x.fillna(x.median())).reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "df[\"Max_VisibilityKm\"] = df.groupby(\"Region\")[\"Max_VisibilityKm\"].transform(lambda x: x.fillna(x.median())).reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "df[\"Min_VisibilitykM\"] = df.groupby(\"Region\")[\"Min_VisibilitykM\"].transform(lambda x: x.fillna(x.median())).reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "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>StoreID</th>\n",
       "      <th>Date</th>\n",
       "      <th>IsHoliday</th>\n",
       "      <th>IsOpen</th>\n",
       "      <th>HasPromotions</th>\n",
       "      <th>StoreType</th>\n",
       "      <th>AssortmentType</th>\n",
       "      <th>NearestCompetitor</th>\n",
       "      <th>Region</th>\n",
       "      <th>Region_AreaKM2</th>\n",
       "      <th>...</th>\n",
       "      <th>Min_Sea_Level_PressurehPa</th>\n",
       "      <th>Min_TemperatureC</th>\n",
       "      <th>Min_VisibilitykM</th>\n",
       "      <th>Precipitationmm</th>\n",
       "      <th>WindDirDegrees</th>\n",
       "      <th>D_Day</th>\n",
       "      <th>D_DayOfYear</th>\n",
       "      <th>D_Month</th>\n",
       "      <th>D_Year</th>\n",
       "      <th>D_DayOfweek</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1000</td>\n",
       "      <td>01/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1011</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>180</td>\n",
       "      <td>1</td>\n",
       "      <td>60</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1000</td>\n",
       "      <td>02/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1009</td>\n",
       "      <td>3</td>\n",
       "      <td>1.0</td>\n",
       "      <td>5.08</td>\n",
       "      <td>315</td>\n",
       "      <td>2</td>\n",
       "      <td>61</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1000</td>\n",
       "      <td>03/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1013</td>\n",
       "      <td>-2</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>210</td>\n",
       "      <td>3</td>\n",
       "      <td>62</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1000</td>\n",
       "      <td>04/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1002</td>\n",
       "      <td>1</td>\n",
       "      <td>3.0</td>\n",
       "      <td>3.05</td>\n",
       "      <td>193</td>\n",
       "      <td>4</td>\n",
       "      <td>63</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1000</td>\n",
       "      <td>05/03/2018</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Hyper Market</td>\n",
       "      <td>General</td>\n",
       "      <td>326</td>\n",
       "      <td>7</td>\n",
       "      <td>9643</td>\n",
       "      <td>...</td>\n",
       "      <td>1000</td>\n",
       "      <td>2</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.25</td>\n",
       "      <td>247</td>\n",
       "      <td>5</td>\n",
       "      <td>64</td>\n",
       "      <td>3</td>\n",
       "      <td>2018</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 38 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   StoreID        Date  IsHoliday  IsOpen  HasPromotions     StoreType  \\\n",
       "0     1000  01/03/2018          0       1              0  Hyper Market   \n",
       "1     1000  02/03/2018          0       1              0  Hyper Market   \n",
       "2     1000  03/03/2018          0       1              0  Hyper Market   \n",
       "3     1000  04/03/2018          0       0              0  Hyper Market   \n",
       "4     1000  05/03/2018          0       1              1  Hyper Market   \n",
       "\n",
       "  AssortmentType  NearestCompetitor  Region  Region_AreaKM2     ...       \\\n",
       "0        General                326       7            9643     ...        \n",
       "1        General                326       7            9643     ...        \n",
       "2        General                326       7            9643     ...        \n",
       "3        General                326       7            9643     ...        \n",
       "4        General                326       7            9643     ...        \n",
       "\n",
       "   Min_Sea_Level_PressurehPa  Min_TemperatureC  Min_VisibilitykM  \\\n",
       "0                       1011                 2               0.0   \n",
       "1                       1009                 3               1.0   \n",
       "2                       1013                -2               2.0   \n",
       "3                       1002                 1               3.0   \n",
       "4                       1000                 2               4.0   \n",
       "\n",
       "  Precipitationmm  WindDirDegrees  D_Day  D_DayOfYear  D_Month  D_Year  \\\n",
       "0            0.00             180      1           60        3    2018   \n",
       "1            5.08             315      2           61        3    2018   \n",
       "2            0.00             210      3           62        3    2018   \n",
       "3            3.05             193      4           63        3    2018   \n",
       "4            0.25             247      5           64        3    2018   \n",
       "\n",
       "   D_DayOfweek  \n",
       "0            3  \n",
       "1            4  \n",
       "2            5  \n",
       "3            6  \n",
       "4            0  \n",
       "\n",
       "[5 rows x 38 columns]"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Final check\n",
    "\n",
    "Let's check again how much are the remaining missing ratios"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "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 Ratio</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [Missing Ratio]\n",
       "Index: []"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_na = (df.isnull().sum() / len(df)) * 100\n",
    "df_na = df_na.drop(df_na[df_na == 0].index).sort_values(ascending=False)[:30]\n",
    "missing_data = pd.DataFrame({'Missing Ratio': df_na})\n",
    "missing_data.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Write to file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "# save the imputed dataset\n",
    "df.to_csv('./dataset/preprocessed1_imputation_' + work_on + '.csv', index=False)"
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}