Predicting sales EQ(interupt01h).ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from statsmodels.graphics.tsaplots import plot_acf\n",
    "import seaborn as sns\n",
    "import pickle\n",
    "\n",
    "%matplotlib inline\n",
    "train_data_org = pd.read_excel('data/Training-Data-Sets.xlsx')\n",
    "test_data_org = pd.read_excel('data/Test dataset v1.xlsx')"
   ]
  },
  {
   "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>Day</th>\n",
       "      <th>EQ</th>\n",
       "      <th>Social_Search_Impressions</th>\n",
       "      <th>Social_Search_Working_cost</th>\n",
       "      <th>Digital_Impressions</th>\n",
       "      <th>Digital_Working_cost</th>\n",
       "      <th>Print_Impressions.Ads40</th>\n",
       "      <th>Print_Working_Cost.Ads50</th>\n",
       "      <th>OOH_Impressions</th>\n",
       "      <th>OOH_Working_Cost</th>\n",
       "      <th>...</th>\n",
       "      <th>Magazine_Impressions_pct</th>\n",
       "      <th>TV_GRP</th>\n",
       "      <th>Competitor1_RPI</th>\n",
       "      <th>Competitor2_RPI</th>\n",
       "      <th>Competitor3_RPI</th>\n",
       "      <th>Competitor4_RPI</th>\n",
       "      <th>EQ_Category</th>\n",
       "      <th>EQ_Subcategory</th>\n",
       "      <th>pct_PromoMarketDollars_Subcategory</th>\n",
       "      <th>RPI_Subcategory</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>718.458667</td>\n",
       "      <td>22256928</td>\n",
       "      <td>56812</td>\n",
       "      <td>7724107</td>\n",
       "      <td>238700</td>\n",
       "      <td>151438</td>\n",
       "      <td>1044</td>\n",
       "      <td>112229070</td>\n",
       "      <td>2133614</td>\n",
       "      <td>...</td>\n",
       "      <td>54.17</td>\n",
       "      <td>16.46</td>\n",
       "      <td>105.71</td>\n",
       "      <td>36.13</td>\n",
       "      <td>46.36</td>\n",
       "      <td>71.83</td>\n",
       "      <td>5420048</td>\n",
       "      <td>475559</td>\n",
       "      <td>0.3766</td>\n",
       "      <td>45.77</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>25.521137</td>\n",
       "      <td>4239408</td>\n",
       "      <td>105695</td>\n",
       "      <td>5844288</td>\n",
       "      <td>188902</td>\n",
       "      <td>264008</td>\n",
       "      <td>113582</td>\n",
       "      <td>284597324</td>\n",
       "      <td>1719318</td>\n",
       "      <td>...</td>\n",
       "      <td>65.79</td>\n",
       "      <td>15.64</td>\n",
       "      <td>111.79</td>\n",
       "      <td>43.31</td>\n",
       "      <td>47.58</td>\n",
       "      <td>67.90</td>\n",
       "      <td>12155631</td>\n",
       "      <td>371540</td>\n",
       "      <td>0.2515</td>\n",
       "      <td>35.02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>268.288044</td>\n",
       "      <td>6708500</td>\n",
       "      <td>87686</td>\n",
       "      <td>13008485</td>\n",
       "      <td>19704</td>\n",
       "      <td>150505</td>\n",
       "      <td>38501</td>\n",
       "      <td>807781054</td>\n",
       "      <td>1569740</td>\n",
       "      <td>...</td>\n",
       "      <td>45.08</td>\n",
       "      <td>23.91</td>\n",
       "      <td>110.15</td>\n",
       "      <td>38.26</td>\n",
       "      <td>49.26</td>\n",
       "      <td>72.70</td>\n",
       "      <td>11939870</td>\n",
       "      <td>225984</td>\n",
       "      <td>0.3679</td>\n",
       "      <td>46.22</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>209.071599</td>\n",
       "      <td>36835247</td>\n",
       "      <td>70791</td>\n",
       "      <td>2520814</td>\n",
       "      <td>200111</td>\n",
       "      <td>253458</td>\n",
       "      <td>53719</td>\n",
       "      <td>667441334</td>\n",
       "      <td>922723</td>\n",
       "      <td>...</td>\n",
       "      <td>75.24</td>\n",
       "      <td>13.06</td>\n",
       "      <td>116.74</td>\n",
       "      <td>39.37</td>\n",
       "      <td>44.21</td>\n",
       "      <td>73.01</td>\n",
       "      <td>7045541</td>\n",
       "      <td>551342</td>\n",
       "      <td>0.0504</td>\n",
       "      <td>38.10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>3482.169185</td>\n",
       "      <td>23693467</td>\n",
       "      <td>75610</td>\n",
       "      <td>9276779</td>\n",
       "      <td>65532</td>\n",
       "      <td>278877</td>\n",
       "      <td>95178</td>\n",
       "      <td>74045467</td>\n",
       "      <td>1834970</td>\n",
       "      <td>...</td>\n",
       "      <td>56.85</td>\n",
       "      <td>40.72</td>\n",
       "      <td>115.18</td>\n",
       "      <td>36.85</td>\n",
       "      <td>45.46</td>\n",
       "      <td>75.84</td>\n",
       "      <td>11488805</td>\n",
       "      <td>254143</td>\n",
       "      <td>0.2219</td>\n",
       "      <td>39.16</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 39 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Day           EQ  Social_Search_Impressions  Social_Search_Working_cost  \\\n",
       "0    1   718.458667                   22256928                       56812   \n",
       "1    2    25.521137                    4239408                      105695   \n",
       "2    3   268.288044                    6708500                       87686   \n",
       "3    4   209.071599                   36835247                       70791   \n",
       "4    5  3482.169185                   23693467                       75610   \n",
       "\n",
       "   Digital_Impressions  Digital_Working_cost  Print_Impressions.Ads40  \\\n",
       "0              7724107                238700                   151438   \n",
       "1              5844288                188902                   264008   \n",
       "2             13008485                 19704                   150505   \n",
       "3              2520814                200111                   253458   \n",
       "4              9276779                 65532                   278877   \n",
       "\n",
       "   Print_Working_Cost.Ads50  OOH_Impressions  OOH_Working_Cost  ...  \\\n",
       "0                      1044        112229070           2133614  ...   \n",
       "1                    113582        284597324           1719318  ...   \n",
       "2                     38501        807781054           1569740  ...   \n",
       "3                     53719        667441334            922723  ...   \n",
       "4                     95178         74045467           1834970  ...   \n",
       "\n",
       "   Magazine_Impressions_pct  TV_GRP  Competitor1_RPI  Competitor2_RPI  \\\n",
       "0                     54.17   16.46           105.71            36.13   \n",
       "1                     65.79   15.64           111.79            43.31   \n",
       "2                     45.08   23.91           110.15            38.26   \n",
       "3                     75.24   13.06           116.74            39.37   \n",
       "4                     56.85   40.72           115.18            36.85   \n",
       "\n",
       "   Competitor3_RPI  Competitor4_RPI  EQ_Category  EQ_Subcategory  \\\n",
       "0            46.36            71.83      5420048          475559   \n",
       "1            47.58            67.90     12155631          371540   \n",
       "2            49.26            72.70     11939870          225984   \n",
       "3            44.21            73.01      7045541          551342   \n",
       "4            45.46            75.84     11488805          254143   \n",
       "\n",
       "   pct_PromoMarketDollars_Subcategory  RPI_Subcategory  \n",
       "0                              0.3766            45.77  \n",
       "1                              0.2515            35.02  \n",
       "2                              0.3679            46.22  \n",
       "3                              0.0504            38.10  \n",
       "4                              0.2219            39.16  \n",
       "\n",
       "[5 rows x 39 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data_org.head()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>Day</th>\n",
       "      <th>EQ</th>\n",
       "      <th>Social_Search_Impressions</th>\n",
       "      <th>Social_Search_Working_cost</th>\n",
       "      <th>Digital_Impressions</th>\n",
       "      <th>Digital_Working_cost</th>\n",
       "      <th>Print_Impressions.Ads40</th>\n",
       "      <th>Print_Working_Cost.Ads50</th>\n",
       "      <th>OOH_Impressions</th>\n",
       "      <th>OOH_Working_Cost</th>\n",
       "      <th>...</th>\n",
       "      <th>Magazine_Impressions_pct</th>\n",
       "      <th>TV_GRP</th>\n",
       "      <th>Competitor1_RPI</th>\n",
       "      <th>Competitor2_RPI</th>\n",
       "      <th>Competitor3_RPI</th>\n",
       "      <th>Competitor4_RPI</th>\n",
       "      <th>EQ_Category</th>\n",
       "      <th>EQ_Subcategory</th>\n",
       "      <th>pct_PromoMarketDollars_Subcategory</th>\n",
       "      <th>RPI_Subcategory</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>12000.00000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>1.200000e+04</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>1.200000e+04</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>1.200000e+04</td>\n",
       "      <td>1.200000e+04</td>\n",
       "      <td>...</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>1.200000e+04</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "      <td>12000.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>6000.50000</td>\n",
       "      <td>638.008496</td>\n",
       "      <td>1.962056e+07</td>\n",
       "      <td>63131.860167</td>\n",
       "      <td>6.663405e+06</td>\n",
       "      <td>218973.240167</td>\n",
       "      <td>194404.538917</td>\n",
       "      <td>95406.492167</td>\n",
       "      <td>4.526812e+08</td>\n",
       "      <td>1.975918e+06</td>\n",
       "      <td>...</td>\n",
       "      <td>50.354613</td>\n",
       "      <td>28.902704</td>\n",
       "      <td>106.173423</td>\n",
       "      <td>38.655960</td>\n",
       "      <td>46.162093</td>\n",
       "      <td>70.641081</td>\n",
       "      <td>6.688170e+06</td>\n",
       "      <td>549685.650917</td>\n",
       "      <td>0.218851</td>\n",
       "      <td>40.159309</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>3464.24595</td>\n",
       "      <td>1200.109675</td>\n",
       "      <td>1.084298e+07</td>\n",
       "      <td>34518.859817</td>\n",
       "      <td>3.824995e+06</td>\n",
       "      <td>124075.882373</td>\n",
       "      <td>85857.419707</td>\n",
       "      <td>55041.846012</td>\n",
       "      <td>2.308381e+08</td>\n",
       "      <td>1.017836e+06</td>\n",
       "      <td>...</td>\n",
       "      <td>16.440054</td>\n",
       "      <td>9.506248</td>\n",
       "      <td>8.229686</td>\n",
       "      <td>3.903151</td>\n",
       "      <td>1.893655</td>\n",
       "      <td>5.081959</td>\n",
       "      <td>3.179592e+06</td>\n",
       "      <td>197373.143664</td>\n",
       "      <td>0.098834</td>\n",
       "      <td>5.129030</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1.00000</td>\n",
       "      <td>0.019459</td>\n",
       "      <td>8.741110e+05</td>\n",
       "      <td>3546.000000</td>\n",
       "      <td>2.344000e+04</td>\n",
       "      <td>3493.000000</td>\n",
       "      <td>46372.000000</td>\n",
       "      <td>462.000000</td>\n",
       "      <td>5.435061e+07</td>\n",
       "      <td>2.374290e+05</td>\n",
       "      <td>...</td>\n",
       "      <td>21.890000</td>\n",
       "      <td>12.340000</td>\n",
       "      <td>91.900000</td>\n",
       "      <td>31.900000</td>\n",
       "      <td>42.900000</td>\n",
       "      <td>61.900000</td>\n",
       "      <td>1.234920e+06</td>\n",
       "      <td>209473.000000</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>31.230000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>3000.75000</td>\n",
       "      <td>57.604317</td>\n",
       "      <td>1.021389e+07</td>\n",
       "      <td>33164.000000</td>\n",
       "      <td>3.330268e+06</td>\n",
       "      <td>112318.250000</td>\n",
       "      <td>120124.750000</td>\n",
       "      <td>47610.000000</td>\n",
       "      <td>2.519760e+08</td>\n",
       "      <td>1.095344e+06</td>\n",
       "      <td>...</td>\n",
       "      <td>36.100000</td>\n",
       "      <td>20.720000</td>\n",
       "      <td>99.030000</td>\n",
       "      <td>35.260000</td>\n",
       "      <td>44.520000</td>\n",
       "      <td>66.290000</td>\n",
       "      <td>3.965402e+06</td>\n",
       "      <td>380095.250000</td>\n",
       "      <td>0.133175</td>\n",
       "      <td>35.730000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>6000.50000</td>\n",
       "      <td>210.732487</td>\n",
       "      <td>1.949458e+07</td>\n",
       "      <td>62888.500000</td>\n",
       "      <td>6.715113e+06</td>\n",
       "      <td>218230.500000</td>\n",
       "      <td>193609.500000</td>\n",
       "      <td>95586.000000</td>\n",
       "      <td>4.540579e+08</td>\n",
       "      <td>1.959212e+06</td>\n",
       "      <td>...</td>\n",
       "      <td>50.190000</td>\n",
       "      <td>28.960000</td>\n",
       "      <td>106.150000</td>\n",
       "      <td>38.625000</td>\n",
       "      <td>46.170000</td>\n",
       "      <td>70.670000</td>\n",
       "      <td>6.619380e+06</td>\n",
       "      <td>547465.000000</td>\n",
       "      <td>0.220000</td>\n",
       "      <td>40.200000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>9000.25000</td>\n",
       "      <td>665.093566</td>\n",
       "      <td>2.913852e+07</td>\n",
       "      <td>92461.750000</td>\n",
       "      <td>9.956033e+06</td>\n",
       "      <td>326631.250000</td>\n",
       "      <td>268843.750000</td>\n",
       "      <td>143789.750000</td>\n",
       "      <td>6.557911e+08</td>\n",
       "      <td>2.854418e+06</td>\n",
       "      <td>...</td>\n",
       "      <td>64.572500</td>\n",
       "      <td>37.240000</td>\n",
       "      <td>113.282500</td>\n",
       "      <td>42.030000</td>\n",
       "      <td>47.790000</td>\n",
       "      <td>75.030000</td>\n",
       "      <td>9.430345e+06</td>\n",
       "      <td>719804.750000</td>\n",
       "      <td>0.305400</td>\n",
       "      <td>44.560000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>12000.00000</td>\n",
       "      <td>18557.563795</td>\n",
       "      <td>3.827240e+07</td>\n",
       "      <td>123421.000000</td>\n",
       "      <td>1.323874e+07</td>\n",
       "      <td>432340.000000</td>\n",
       "      <td>342242.000000</td>\n",
       "      <td>190389.000000</td>\n",
       "      <td>8.493609e+08</td>\n",
       "      <td>3.748194e+06</td>\n",
       "      <td>...</td>\n",
       "      <td>78.730000</td>\n",
       "      <td>45.340000</td>\n",
       "      <td>120.440000</td>\n",
       "      <td>45.440000</td>\n",
       "      <td>49.440000</td>\n",
       "      <td>79.440000</td>\n",
       "      <td>1.223400e+07</td>\n",
       "      <td>893820.000000</td>\n",
       "      <td>0.389300</td>\n",
       "      <td>49.020000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8 rows × 39 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               Day            EQ  Social_Search_Impressions  \\\n",
       "count  12000.00000  12000.000000               1.200000e+04   \n",
       "mean    6000.50000    638.008496               1.962056e+07   \n",
       "std     3464.24595   1200.109675               1.084298e+07   \n",
       "min        1.00000      0.019459               8.741110e+05   \n",
       "25%     3000.75000     57.604317               1.021389e+07   \n",
       "50%     6000.50000    210.732487               1.949458e+07   \n",
       "75%     9000.25000    665.093566               2.913852e+07   \n",
       "max    12000.00000  18557.563795               3.827240e+07   \n",
       "\n",
       "       Social_Search_Working_cost  Digital_Impressions  Digital_Working_cost  \\\n",
       "count                12000.000000         1.200000e+04          12000.000000   \n",
       "mean                 63131.860167         6.663405e+06         218973.240167   \n",
       "std                  34518.859817         3.824995e+06         124075.882373   \n",
       "min                   3546.000000         2.344000e+04           3493.000000   \n",
       "25%                  33164.000000         3.330268e+06         112318.250000   \n",
       "50%                  62888.500000         6.715113e+06         218230.500000   \n",
       "75%                  92461.750000         9.956033e+06         326631.250000   \n",
       "max                 123421.000000         1.323874e+07         432340.000000   \n",
       "\n",
       "       Print_Impressions.Ads40  Print_Working_Cost.Ads50  OOH_Impressions  \\\n",
       "count             12000.000000              12000.000000     1.200000e+04   \n",
       "mean             194404.538917              95406.492167     4.526812e+08   \n",
       "std               85857.419707              55041.846012     2.308381e+08   \n",
       "min               46372.000000                462.000000     5.435061e+07   \n",
       "25%              120124.750000              47610.000000     2.519760e+08   \n",
       "50%              193609.500000              95586.000000     4.540579e+08   \n",
       "75%              268843.750000             143789.750000     6.557911e+08   \n",
       "max              342242.000000             190389.000000     8.493609e+08   \n",
       "\n",
       "       OOH_Working_Cost  ...  Magazine_Impressions_pct        TV_GRP  \\\n",
       "count      1.200000e+04  ...              12000.000000  12000.000000   \n",
       "mean       1.975918e+06  ...                 50.354613     28.902704   \n",
       "std        1.017836e+06  ...                 16.440054      9.506248   \n",
       "min        2.374290e+05  ...                 21.890000     12.340000   \n",
       "25%        1.095344e+06  ...                 36.100000     20.720000   \n",
       "50%        1.959212e+06  ...                 50.190000     28.960000   \n",
       "75%        2.854418e+06  ...                 64.572500     37.240000   \n",
       "max        3.748194e+06  ...                 78.730000     45.340000   \n",
       "\n",
       "       Competitor1_RPI  Competitor2_RPI  Competitor3_RPI  Competitor4_RPI  \\\n",
       "count     12000.000000     12000.000000     12000.000000     12000.000000   \n",
       "mean        106.173423        38.655960        46.162093        70.641081   \n",
       "std           8.229686         3.903151         1.893655         5.081959   \n",
       "min          91.900000        31.900000        42.900000        61.900000   \n",
       "25%          99.030000        35.260000        44.520000        66.290000   \n",
       "50%         106.150000        38.625000        46.170000        70.670000   \n",
       "75%         113.282500        42.030000        47.790000        75.030000   \n",
       "max         120.440000        45.440000        49.440000        79.440000   \n",
       "\n",
       "        EQ_Category  EQ_Subcategory  pct_PromoMarketDollars_Subcategory  \\\n",
       "count  1.200000e+04    12000.000000                        12000.000000   \n",
       "mean   6.688170e+06   549685.650917                            0.218851   \n",
       "std    3.179592e+06   197373.143664                            0.098834   \n",
       "min    1.234920e+06   209473.000000                            0.047200   \n",
       "25%    3.965402e+06   380095.250000                            0.133175   \n",
       "50%    6.619380e+06   547465.000000                            0.220000   \n",
       "75%    9.430345e+06   719804.750000                            0.305400   \n",
       "max    1.223400e+07   893820.000000                            0.389300   \n",
       "\n",
       "       RPI_Subcategory  \n",
       "count     12000.000000  \n",
       "mean         40.159309  \n",
       "std           5.129030  \n",
       "min          31.230000  \n",
       "25%          35.730000  \n",
       "50%          40.200000  \n",
       "75%          44.560000  \n",
       "max          49.020000  \n",
       "\n",
       "[8 rows x 39 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data_org.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Autocorrelation\n",
    "Because we are working with timeseries data, we need to know if the target variable is autocorrelated."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "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": [
    "train_data_y = train_data_org['EQ']\n",
    "plot_acf(train_data_y[:1000])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There seems to be no autocorrelation present. The series also seems to stationary, considering how quickly the plot falls to zero.\n",
    "\n",
    "Since we how to forecast sales for future time periods, we can create a lag feature EQ_next with the next time period's EQ."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>Day</th>\n",
       "      <th>EQ</th>\n",
       "      <th>Social_Search_Impressions</th>\n",
       "      <th>Social_Search_Working_cost</th>\n",
       "      <th>Digital_Impressions</th>\n",
       "      <th>Digital_Working_cost</th>\n",
       "      <th>Print_Impressions.Ads40</th>\n",
       "      <th>Print_Working_Cost.Ads50</th>\n",
       "      <th>OOH_Impressions</th>\n",
       "      <th>OOH_Working_Cost</th>\n",
       "      <th>...</th>\n",
       "      <th>TV_GRP</th>\n",
       "      <th>Competitor1_RPI</th>\n",
       "      <th>Competitor2_RPI</th>\n",
       "      <th>Competitor3_RPI</th>\n",
       "      <th>Competitor4_RPI</th>\n",
       "      <th>EQ_Category</th>\n",
       "      <th>EQ_Subcategory</th>\n",
       "      <th>pct_PromoMarketDollars_Subcategory</th>\n",
       "      <th>RPI_Subcategory</th>\n",
       "      <th>EQ_next</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>25.521137</td>\n",
       "      <td>4239408</td>\n",
       "      <td>105695</td>\n",
       "      <td>5844288</td>\n",
       "      <td>188902</td>\n",
       "      <td>264008</td>\n",
       "      <td>113582</td>\n",
       "      <td>284597324</td>\n",
       "      <td>1719318</td>\n",
       "      <td>...</td>\n",
       "      <td>15.64</td>\n",
       "      <td>111.79</td>\n",
       "      <td>43.31</td>\n",
       "      <td>47.58</td>\n",
       "      <td>67.90</td>\n",
       "      <td>12155631</td>\n",
       "      <td>371540</td>\n",
       "      <td>0.2515</td>\n",
       "      <td>35.02</td>\n",
       "      <td>718.458667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>268.288044</td>\n",
       "      <td>6708500</td>\n",
       "      <td>87686</td>\n",
       "      <td>13008485</td>\n",
       "      <td>19704</td>\n",
       "      <td>150505</td>\n",
       "      <td>38501</td>\n",
       "      <td>807781054</td>\n",
       "      <td>1569740</td>\n",
       "      <td>...</td>\n",
       "      <td>23.91</td>\n",
       "      <td>110.15</td>\n",
       "      <td>38.26</td>\n",
       "      <td>49.26</td>\n",
       "      <td>72.70</td>\n",
       "      <td>11939870</td>\n",
       "      <td>225984</td>\n",
       "      <td>0.3679</td>\n",
       "      <td>46.22</td>\n",
       "      <td>25.521137</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>209.071599</td>\n",
       "      <td>36835247</td>\n",
       "      <td>70791</td>\n",
       "      <td>2520814</td>\n",
       "      <td>200111</td>\n",
       "      <td>253458</td>\n",
       "      <td>53719</td>\n",
       "      <td>667441334</td>\n",
       "      <td>922723</td>\n",
       "      <td>...</td>\n",
       "      <td>13.06</td>\n",
       "      <td>116.74</td>\n",
       "      <td>39.37</td>\n",
       "      <td>44.21</td>\n",
       "      <td>73.01</td>\n",
       "      <td>7045541</td>\n",
       "      <td>551342</td>\n",
       "      <td>0.0504</td>\n",
       "      <td>38.10</td>\n",
       "      <td>268.288044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>3482.169185</td>\n",
       "      <td>23693467</td>\n",
       "      <td>75610</td>\n",
       "      <td>9276779</td>\n",
       "      <td>65532</td>\n",
       "      <td>278877</td>\n",
       "      <td>95178</td>\n",
       "      <td>74045467</td>\n",
       "      <td>1834970</td>\n",
       "      <td>...</td>\n",
       "      <td>40.72</td>\n",
       "      <td>115.18</td>\n",
       "      <td>36.85</td>\n",
       "      <td>45.46</td>\n",
       "      <td>75.84</td>\n",
       "      <td>11488805</td>\n",
       "      <td>254143</td>\n",
       "      <td>0.2219</td>\n",
       "      <td>39.16</td>\n",
       "      <td>209.071599</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>6</td>\n",
       "      <td>55.181083</td>\n",
       "      <td>13925382</td>\n",
       "      <td>114740</td>\n",
       "      <td>2733356</td>\n",
       "      <td>264751</td>\n",
       "      <td>112292</td>\n",
       "      <td>145635</td>\n",
       "      <td>88407557</td>\n",
       "      <td>2864910</td>\n",
       "      <td>...</td>\n",
       "      <td>17.91</td>\n",
       "      <td>118.39</td>\n",
       "      <td>35.87</td>\n",
       "      <td>42.90</td>\n",
       "      <td>66.35</td>\n",
       "      <td>7909917</td>\n",
       "      <td>307429</td>\n",
       "      <td>0.1996</td>\n",
       "      <td>39.26</td>\n",
       "      <td>3482.169185</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 40 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Day           EQ  Social_Search_Impressions  Social_Search_Working_cost  \\\n",
       "1    2    25.521137                    4239408                      105695   \n",
       "2    3   268.288044                    6708500                       87686   \n",
       "3    4   209.071599                   36835247                       70791   \n",
       "4    5  3482.169185                   23693467                       75610   \n",
       "5    6    55.181083                   13925382                      114740   \n",
       "\n",
       "   Digital_Impressions  Digital_Working_cost  Print_Impressions.Ads40  \\\n",
       "1              5844288                188902                   264008   \n",
       "2             13008485                 19704                   150505   \n",
       "3              2520814                200111                   253458   \n",
       "4              9276779                 65532                   278877   \n",
       "5              2733356                264751                   112292   \n",
       "\n",
       "   Print_Working_Cost.Ads50  OOH_Impressions  OOH_Working_Cost  ...  TV_GRP  \\\n",
       "1                    113582        284597324           1719318  ...   15.64   \n",
       "2                     38501        807781054           1569740  ...   23.91   \n",
       "3                     53719        667441334            922723  ...   13.06   \n",
       "4                     95178         74045467           1834970  ...   40.72   \n",
       "5                    145635         88407557           2864910  ...   17.91   \n",
       "\n",
       "   Competitor1_RPI  Competitor2_RPI  Competitor3_RPI  Competitor4_RPI  \\\n",
       "1           111.79            43.31            47.58            67.90   \n",
       "2           110.15            38.26            49.26            72.70   \n",
       "3           116.74            39.37            44.21            73.01   \n",
       "4           115.18            36.85            45.46            75.84   \n",
       "5           118.39            35.87            42.90            66.35   \n",
       "\n",
       "   EQ_Category  EQ_Subcategory  pct_PromoMarketDollars_Subcategory  \\\n",
       "1     12155631          371540                              0.2515   \n",
       "2     11939870          225984                              0.3679   \n",
       "3      7045541          551342                              0.0504   \n",
       "4     11488805          254143                              0.2219   \n",
       "5      7909917          307429                              0.1996   \n",
       "\n",
       "   RPI_Subcategory      EQ_next  \n",
       "1            35.02   718.458667  \n",
       "2            46.22    25.521137  \n",
       "3            38.10   268.288044  \n",
       "4            39.16   209.071599  \n",
       "5            39.26  3482.169185  \n",
       "\n",
       "[5 rows x 40 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def transform_ts_supervised(dataframe):\n",
    "    eq = dataframe['EQ']\n",
    "    \n",
    "    eq_next = eq.shift(periods=1)\n",
    "    dataframe['EQ_next']=list(eq_next)\n",
    "    dataframe=dataframe.dropna()\n",
    "    return dataframe\n",
    "\n",
    "train_data = transform_ts_supervised(train_data_org)\n",
    "train_data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature selection\n",
    "### 1. Removing correlated variables\n",
    "If features are highly correlated, then we have redundant information, thus features highly correlated with another can be removed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f0164b8f810>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "data_corr = train_data.drop(['Day'], axis=1)\n",
    "corr = data_corr.corr()\n",
    "sns.heatmap(corr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Seems like there is little no correlation between them.\n",
    "\n",
    "### 2. Manual feature selection/extraction\n",
    "Some features like rainfall, temperature don't affect sales a lot, they can be removed. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = train_data.drop(['Median_Rainfall', 'Median_Temp'], axis=1)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can create a new feature Avg_Comp_per which is the average relative performance of all 4 competitors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>Day</th>\n",
       "      <th>EQ</th>\n",
       "      <th>Social_Search_Impressions</th>\n",
       "      <th>Social_Search_Working_cost</th>\n",
       "      <th>Digital_Impressions</th>\n",
       "      <th>Digital_Working_cost</th>\n",
       "      <th>Print_Impressions.Ads40</th>\n",
       "      <th>Print_Working_Cost.Ads50</th>\n",
       "      <th>OOH_Impressions</th>\n",
       "      <th>OOH_Working_Cost</th>\n",
       "      <th>...</th>\n",
       "      <th>Competitor1_RPI</th>\n",
       "      <th>Competitor2_RPI</th>\n",
       "      <th>Competitor3_RPI</th>\n",
       "      <th>Competitor4_RPI</th>\n",
       "      <th>EQ_Category</th>\n",
       "      <th>EQ_Subcategory</th>\n",
       "      <th>pct_PromoMarketDollars_Subcategory</th>\n",
       "      <th>RPI_Subcategory</th>\n",
       "      <th>EQ_next</th>\n",
       "      <th>Avg_Comp_Per</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>25.521137</td>\n",
       "      <td>4239408</td>\n",
       "      <td>105695</td>\n",
       "      <td>5844288</td>\n",
       "      <td>188902</td>\n",
       "      <td>264008</td>\n",
       "      <td>113582</td>\n",
       "      <td>284597324</td>\n",
       "      <td>1719318</td>\n",
       "      <td>...</td>\n",
       "      <td>111.79</td>\n",
       "      <td>43.31</td>\n",
       "      <td>47.58</td>\n",
       "      <td>67.90</td>\n",
       "      <td>12155631</td>\n",
       "      <td>371540</td>\n",
       "      <td>0.2515</td>\n",
       "      <td>35.02</td>\n",
       "      <td>718.458667</td>\n",
       "      <td>67.6450</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>268.288044</td>\n",
       "      <td>6708500</td>\n",
       "      <td>87686</td>\n",
       "      <td>13008485</td>\n",
       "      <td>19704</td>\n",
       "      <td>150505</td>\n",
       "      <td>38501</td>\n",
       "      <td>807781054</td>\n",
       "      <td>1569740</td>\n",
       "      <td>...</td>\n",
       "      <td>110.15</td>\n",
       "      <td>38.26</td>\n",
       "      <td>49.26</td>\n",
       "      <td>72.70</td>\n",
       "      <td>11939870</td>\n",
       "      <td>225984</td>\n",
       "      <td>0.3679</td>\n",
       "      <td>46.22</td>\n",
       "      <td>25.521137</td>\n",
       "      <td>67.5925</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>209.071599</td>\n",
       "      <td>36835247</td>\n",
       "      <td>70791</td>\n",
       "      <td>2520814</td>\n",
       "      <td>200111</td>\n",
       "      <td>253458</td>\n",
       "      <td>53719</td>\n",
       "      <td>667441334</td>\n",
       "      <td>922723</td>\n",
       "      <td>...</td>\n",
       "      <td>116.74</td>\n",
       "      <td>39.37</td>\n",
       "      <td>44.21</td>\n",
       "      <td>73.01</td>\n",
       "      <td>7045541</td>\n",
       "      <td>551342</td>\n",
       "      <td>0.0504</td>\n",
       "      <td>38.10</td>\n",
       "      <td>268.288044</td>\n",
       "      <td>68.3325</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>3482.169185</td>\n",
       "      <td>23693467</td>\n",
       "      <td>75610</td>\n",
       "      <td>9276779</td>\n",
       "      <td>65532</td>\n",
       "      <td>278877</td>\n",
       "      <td>95178</td>\n",
       "      <td>74045467</td>\n",
       "      <td>1834970</td>\n",
       "      <td>...</td>\n",
       "      <td>115.18</td>\n",
       "      <td>36.85</td>\n",
       "      <td>45.46</td>\n",
       "      <td>75.84</td>\n",
       "      <td>11488805</td>\n",
       "      <td>254143</td>\n",
       "      <td>0.2219</td>\n",
       "      <td>39.16</td>\n",
       "      <td>209.071599</td>\n",
       "      <td>68.3325</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>6</td>\n",
       "      <td>55.181083</td>\n",
       "      <td>13925382</td>\n",
       "      <td>114740</td>\n",
       "      <td>2733356</td>\n",
       "      <td>264751</td>\n",
       "      <td>112292</td>\n",
       "      <td>145635</td>\n",
       "      <td>88407557</td>\n",
       "      <td>2864910</td>\n",
       "      <td>...</td>\n",
       "      <td>118.39</td>\n",
       "      <td>35.87</td>\n",
       "      <td>42.90</td>\n",
       "      <td>66.35</td>\n",
       "      <td>7909917</td>\n",
       "      <td>307429</td>\n",
       "      <td>0.1996</td>\n",
       "      <td>39.26</td>\n",
       "      <td>3482.169185</td>\n",
       "      <td>65.8775</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 39 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Day           EQ  Social_Search_Impressions  Social_Search_Working_cost  \\\n",
       "1    2    25.521137                    4239408                      105695   \n",
       "2    3   268.288044                    6708500                       87686   \n",
       "3    4   209.071599                   36835247                       70791   \n",
       "4    5  3482.169185                   23693467                       75610   \n",
       "5    6    55.181083                   13925382                      114740   \n",
       "\n",
       "   Digital_Impressions  Digital_Working_cost  Print_Impressions.Ads40  \\\n",
       "1              5844288                188902                   264008   \n",
       "2             13008485                 19704                   150505   \n",
       "3              2520814                200111                   253458   \n",
       "4              9276779                 65532                   278877   \n",
       "5              2733356                264751                   112292   \n",
       "\n",
       "   Print_Working_Cost.Ads50  OOH_Impressions  OOH_Working_Cost  ...  \\\n",
       "1                    113582        284597324           1719318  ...   \n",
       "2                     38501        807781054           1569740  ...   \n",
       "3                     53719        667441334            922723  ...   \n",
       "4                     95178         74045467           1834970  ...   \n",
       "5                    145635         88407557           2864910  ...   \n",
       "\n",
       "   Competitor1_RPI  Competitor2_RPI  Competitor3_RPI  Competitor4_RPI  \\\n",
       "1           111.79            43.31            47.58            67.90   \n",
       "2           110.15            38.26            49.26            72.70   \n",
       "3           116.74            39.37            44.21            73.01   \n",
       "4           115.18            36.85            45.46            75.84   \n",
       "5           118.39            35.87            42.90            66.35   \n",
       "\n",
       "   EQ_Category  EQ_Subcategory  pct_PromoMarketDollars_Subcategory  \\\n",
       "1     12155631          371540                              0.2515   \n",
       "2     11939870          225984                              0.3679   \n",
       "3      7045541          551342                              0.0504   \n",
       "4     11488805          254143                              0.2219   \n",
       "5      7909917          307429                              0.1996   \n",
       "\n",
       "   RPI_Subcategory      EQ_next  Avg_Comp_Per  \n",
       "1            35.02   718.458667       67.6450  \n",
       "2            46.22    25.521137       67.5925  \n",
       "3            38.10   268.288044       68.3325  \n",
       "4            39.16   209.071599       68.3325  \n",
       "5            39.26  3482.169185       65.8775  \n",
       "\n",
       "[5 rows x 39 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def add_price_diff_feature(df):\n",
    "    df['Avg_Comp_Per'] = (df['Competitor1_RPI']+df['Competitor2_RPI']+df['Competitor3_RPI']+df['Competitor4_RPI'])/4\n",
    "    return df\n",
    "train_data = add_price_diff_feature(train_data)    \n",
    "train_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "37\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "BayesianRidge(alpha_1=1e-06, alpha_2=1e-06, alpha_init=None,\n",
       "              compute_score=False, copy_X=True, fit_intercept=True,\n",
       "              lambda_1=1e-06, lambda_2=10000000000.0, lambda_init=None,\n",
       "              n_iter=300, normalize=False, tol=0.001, verbose=False)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import BayesianRidge\n",
    "\n",
    "data_x = train_data.drop(['Day', 'EQ_next'], axis=1)\n",
    "data_y = train_data['EQ_next']\n",
    "\n",
    "print(len(data_x.columns))\n",
    "model = BayesianRidge(lambda_2=1e10)\n",
    "model.fit(data_x, data_y)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean absolute percentage error: 115.64623702700503%\n",
      "[569.38457148 641.51894388 692.27397763 543.82658857 689.37128931\n",
      " 602.12061722 803.97427584 596.64018412 710.08272382 540.80373371\n",
      " 634.0531106  677.37657676 844.77106941 775.90702508 856.16712402\n",
      " 425.37143005 186.26772998 -15.27806131 104.96431625 473.46556839\n",
      " 647.00619954 668.35550006 798.68108485 717.97579659 578.48023989\n",
      " 675.37917479 696.18085752 630.55322021 684.09035366 705.7113558\n",
      " 658.09682851 365.29168307 172.57808448  36.16796015 386.12423551\n",
      " 584.51070815 587.44219609 408.19105168]\n",
      "[504.7849334 490.2264774 479.2446856 489.0574275 477.0319938 487.8552994\n",
      " 466.3992813 546.0530843 464.9256065 357.6486931 298.553287  283.7974301\n",
      " 239.2316222 392.3264313 355.6522644 286.705586  361.4447145 378.2739106\n",
      " 300.9220559 367.5470472 385.53789   332.1503773 237.7135844 193.3007544\n",
      " 179.2925035 173.2373049 247.3154761 284.1832918 274.4307568 205.4999844\n",
      " 250.555064  278.3175236 284.8954797 244.9313589 175.4323186 168.1066651\n",
      " 161.529279  151.6422314]\n"
     ]
    }
   ],
   "source": [
    "# Function to calculate MAPE\n",
    "def mean_absolute_percentage_error(y_true, y_pred): \n",
    "    y_true, y_pred = np.array(y_true), np.array(y_pred)\n",
    "    return np.mean(np.abs((y_true - y_pred) / y_true)) * 100\n",
    "\n",
    "test_data = transform_ts_supervised(test_data_org)\n",
    "test_data = test_data.drop(['Median_Rainfall', 'Median_Temp'], axis=1)\n",
    "\n",
    "test_data = add_price_diff_feature(test_data)\n",
    "\n",
    "test_data_x = test_data.drop(['EQ_next', 'Period'], axis=1)\n",
    "test_data_y = test_data['EQ_next'].values\n",
    "predictions = model.predict(test_data_x)\n",
    "te = mean_absolute_percentage_error(test_data_y, predictions)\n",
    "print('Mean absolute percentage error: {}%'.format(te))\n",
    "pred = np.array(predictions)\n",
    "print(pred)\n",
    "print(np.array(test_data_y))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Machine learning approach"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model name: LinearRegression, CV Mean absolute percentage error: 2985.5549259472346%\n",
      "Model name: SVR, CV Mean absolute percentage error: 965.2153840365361%\n",
      "Model name: RandomForestRegressor, CV Mean absolute percentage error: 3277.6048369342316%\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.svm import SVR, LinearSVR\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.model_selection import KFold\n",
    "\n",
    "models = [LinearRegression(), SVR(), RandomForestRegressor()]\n",
    "for m in models:\n",
    "    score = 0\n",
    "    folds=4\n",
    "    kf = KFold(n_splits=folds)\n",
    "    for train_index, test_index in kf.split(data_x):\n",
    "        X_train, X_test = data_x.iloc[train_index], data_x.iloc[test_index]\n",
    "        y_train, y_test = data_y.iloc[train_index], data_y.iloc[test_index]\n",
    "        m.fit(X_train, y_train)\n",
    "        pred = m.predict(X_test)\n",
    "        te = mean_absolute_percentage_error(y_test, pred)\n",
    "        score+=te\n",
    "    \n",
    "    print('Model name: {}, CV Mean absolute percentage error: {}%'.format(type(m).__name__,score/folds))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Support Vector Machines seem to be pretty good in crossvalidation. Lets train a SVR on the full dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Mean absolute percentage error: 946.2016769326146%\n",
      "Test Mean absolute percentage error: 30.204506282103498%\n"
     ]
    }
   ],
   "source": [
    "model = SVR(C=5e2)\n",
    "model.fit(data_x, data_y)\n",
    "pred_test = model.predict(test_data_x)\n",
    "pred_train = model.predict(data_x)\n",
    "test_error = mean_absolute_percentage_error(test_data_y, pred_test)\n",
    "train_error = mean_absolute_percentage_error(data_y, pred_train)\n",
    "print('Train Mean absolute percentage error: {}%'.format(train_error))\n",
    "print('Test Mean absolute percentage error: {}%'.format(test_error))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Forecasting approach"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have already established that the data has no autocorrelation and the data is stationary, we can fit a ARIMAX model without any need for differencing.\n",
    "\n",
    "Since we are forecasting, there is no question of train or test sets, buts lets fit the algorithm on both sets anyway."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/aftaab/miniconda3/lib/python3.7/site-packages/statsmodels/tsa/base/tsa_model.py:214: ValueWarning: An unsupported index was provided and will be ignored when e.g. forecasting.\n",
      "  ' ignored when e.g. forecasting.', ValueWarning)\n",
      "/home/aftaab/miniconda3/lib/python3.7/site-packages/statsmodels/tsa/base/tsa_model.py:214: ValueWarning: An unsupported index was provided and will be ignored when e.g. forecasting.\n",
      "  ' ignored when e.g. forecasting.', ValueWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                               SARIMAX Results                                \n",
      "==============================================================================\n",
      "Dep. Variable:                     EQ   No. Observations:                11999\n",
      "Model:               SARIMAX(2, 0, 0)   Log Likelihood             -103053.812\n",
      "Date:                Tue, 31 Mar 2020   AIC                         206113.623\n",
      "Time:                        18:07:47   BIC                         206135.800\n",
      "Sample:                             0   HQIC                        206121.063\n",
      "                              - 11999                                         \n",
      "Covariance Type:                  opg                                         \n",
      "==============================================================================\n",
      "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
      "------------------------------------------------------------------------------\n",
      "ar.L1          0.1854      0.008     24.693      0.000       0.171       0.200\n",
      "ar.L2          0.1839      0.007     26.677      0.000       0.170       0.197\n",
      "sigma2      1.693e+06   6057.681    279.491      0.000    1.68e+06     1.7e+06\n",
      "===================================================================================\n",
      "Ljung-Box (Q):                      609.35   Jarque-Bera (JB):            462570.96\n",
      "Prob(Q):                              0.00   Prob(JB):                         0.00\n",
      "Heteroskedasticity (H):               0.82   Skew:                             4.16\n",
      "Prob(H) (two-sided):                  0.00   Kurtosis:                        32.26\n",
      "===================================================================================\n",
      "\n",
      "Warnings:\n",
      "[1] Covariance matrix calculated using the outer product of gradients (complex-step).\n"
     ]
    }
   ],
   "source": [
    "from statsmodels.tsa.statespace.sarimax import SARIMAX\n",
    "train_data = train_data_org.dropna()\n",
    "\n",
    "data_x = train_data.drop(['Median_Rainfall', 'Median_Temp', 'Day', 'EQ'], axis=1)\n",
    "data_x = add_price_diff_feature(data_x)    \n",
    "data_y = train_data['EQ']\n",
    "\n",
    "\n",
    "arimax = SARIMAX(data_y, enforce_invertibility=False, enforce_stationarity=False,\n",
    "                                   order=(2,0,0)).fit()\n",
    "arimax.save(\"forecast_model_train.pkl\")\n",
    "\n",
    "print(arimax.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Lets predict on the train dataset and test sets as well."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Mean absolute percentage error: 1216.3671743772832%\n",
      "test Mean absolute percentage error: 15.69285092931084%\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/aftaab/miniconda3/lib/python3.7/site-packages/statsmodels/tsa/base/tsa_model.py:214: ValueWarning: An unsupported index was provided and will be ignored when e.g. forecasting.\n",
      "  ' ignored when e.g. forecasting.', ValueWarning)\n",
      "/home/aftaab/miniconda3/lib/python3.7/site-packages/statsmodels/tsa/base/tsa_model.py:214: ValueWarning: An unsupported index was provided and will be ignored when e.g. forecasting.\n",
      "  ' ignored when e.g. forecasting.', ValueWarning)\n"
     ]
    },
    {
     "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": [
    "preds = arimax.predict()\n",
    "train_error=mean_absolute_percentage_error(data_y, preds)\n",
    "print('train Mean absolute percentage error: {}%'.format(train_error))\n",
    "\n",
    "test_data = test_data_org.dropna()\n",
    "\n",
    "data_x = test_data.drop(['Median_Rainfall', 'Median_Temp', 'Period', 'EQ'], axis=1)\n",
    "data_x = add_price_diff_feature(data_x)    \n",
    "data_y = test_data['EQ']\n",
    "\n",
    "arimax = SARIMAX(data_y, enforce_invertibility=False, enforce_stationarity=False,\n",
    "                                   order=(2,0,0)).fit()\n",
    "\n",
    "arimax.save(\"forecast_model_test.pkl\")\n",
    "preds = arimax.predict()\n",
    "test_error=mean_absolute_percentage_error(data_y, preds)\n",
    "print('test Mean absolute percentage error: {}%'.format(test_error))\n",
    "\n",
    "\n",
    "plt.plot(data_y)\n",
    "plt.plot(preds, color='red')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Forecasting gives the best test set MAPE by far. Train MAPE is quite large though, but the train set has a large number of outliers which are skewing up the error. Thus this model is the best one so far.\n",
    "\n",
    "Using this model, we can forecast the next 6 periods"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "38    125.571125\n",
      "39    120.403684\n",
      "40    115.448923\n",
      "41    110.698057\n",
      "42    106.142695\n",
      "43    101.774791\n",
      "dtype: float64\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/aftaab/miniconda3/lib/python3.7/site-packages/statsmodels/tsa/base/tsa_model.py:583: ValueWarning: No supported index is available. Prediction results will be given with an integer index beginning at `start`.\n",
      "  ValueWarning)\n"
     ]
    }
   ],
   "source": [
    "# Next 6 values after last test set period.\n",
    "future = arimax.forecast(steps=6)\n",
    "print(future)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Determining most important factors which affect sales\n",
    "Since our forecasting approach can't give us feature importance, we can use one of the machine learning models we made earlier for this purpose(However even though the SVM did best, it cant be used as it used a Gaussian kernel, which makes measuring feature importance impossible). Linear Regression actually did better than Random forests in CV, so lets use that"
   ]
  },
  {
   "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>Feature</th>\n",
       "      <th>Importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>Inflation</td>\n",
       "      <td>197.208069</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>EQ_Base_Price</td>\n",
       "      <td>82.370652</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>Brand_Equity</td>\n",
       "      <td>32.461425</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>pct_PromoMarketDollars_Category</td>\n",
       "      <td>25.480948</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>Any_Disp_pct_ACV</td>\n",
       "      <td>15.741972</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>Fuel_Price</td>\n",
       "      <td>3.958786</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            Feature  Importance\n",
       "15                        Inflation  197.208069\n",
       "22                    EQ_Base_Price   82.370652\n",
       "17                     Brand_Equity   32.461425\n",
       "26  pct_PromoMarketDollars_Category   25.480948\n",
       "21                 Any_Disp_pct_ACV   15.741972\n",
       "14                       Fuel_Price    3.958786"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = LinearRegression()\n",
    "train_data = transform_ts_supervised(train_data_org)\n",
    "data_x = train_data.drop(['Day', 'EQ_next'], axis=1)\n",
    "data_y = train_data['EQ_next']\n",
    "model.fit(data_x, data_y)\n",
    "\n",
    "coefficients = pd.concat([pd.DataFrame(data_x.columns),pd.DataFrame(np.transpose(model.coef_))], axis = 1)\n",
    "coefficients.columns=['Feature', 'Importance']\n",
    "coefficients=coefficients.sort_values(by='Importance', ascending=False)\n",
    "\n",
    "# Showing top 5 positively correlated features\n",
    "coefficients.iloc[:6]"
   ]
  },
  {
   "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>Feature</th>\n",
       "      <th>Importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>pct_PromoMarketDollars_Subcategory</td>\n",
       "      <td>-67.584140</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>Avg_no_of_Items</td>\n",
       "      <td>-7.987799</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>Competitor3_RPI</td>\n",
       "      <td>-6.836056</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>Median_Rainfall</td>\n",
       "      <td>-3.741738</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>RPI_Subcategory</td>\n",
       "      <td>-2.907503</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>RPI_Category</td>\n",
       "      <td>-0.696284</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                               Feature  Importance\n",
       "36  pct_PromoMarketDollars_Subcategory  -67.584140\n",
       "25                     Avg_no_of_Items   -7.987799\n",
       "32                     Competitor3_RPI   -6.836056\n",
       "13                     Median_Rainfall   -3.741738\n",
       "37                     RPI_Subcategory   -2.907503\n",
       "27                        RPI_Category   -0.696284"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Showing top 5 negatively correlated features\n",
    "\n",
    "coefficients = pd.concat([pd.DataFrame(data_x.columns),pd.DataFrame(np.transpose(model.coef_))], axis = 1)\n",
    "coefficients.columns=['Feature', 'Importance']\n",
    "coefficients=coefficients.sort_values(by='Importance')\n",
    "\n",
    "coefficients.iloc[:6]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Loading our saved models and making forecasts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train forecats...\n",
      "11999    52.615896\n",
      "12000    22.813932\n",
      "12001    13.902721\n",
      "12002     6.771580\n",
      "12003     3.811329\n",
      "dtype: float64\n",
      "\n",
      "Test forecasts\n",
      "38    125.571125\n",
      "39    120.403684\n",
      "40    115.448923\n",
      "41    110.698057\n",
      "42    106.142695\n",
      "dtype: float64\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/aftaab/miniconda3/lib/python3.7/site-packages/statsmodels/tsa/base/tsa_model.py:583: ValueWarning: No supported index is available. Prediction results will be given with an integer index beginning at `start`.\n",
      "  ValueWarning)\n",
      "/home/aftaab/miniconda3/lib/python3.7/site-packages/statsmodels/tsa/base/tsa_model.py:583: ValueWarning: No supported index is available. Prediction results will be given with an integer index beginning at `start`.\n",
      "  ValueWarning)\n"
     ]
    }
   ],
   "source": [
    "from statsmodels.tsa.statespace.sarimax import SARIMAXResults\n",
    "\n",
    "forecast_model_train = SARIMAXResults.load('forecast_model_train.pkl')\n",
    "# Make a forecast for the next 6 periods after the last period in the training set\n",
    "future = forecast_model_train.forecast(steps=5)\n",
    "print('Train forecats...')\n",
    "print(future)\n",
    "\n",
    "# Make a forecast for the next 6 periods after the last period in the test set\n",
    "forecast_model_test = SARIMAXResults.load('forecast_model_test.pkl')\n",
    "future = forecast_model_test.forecast(steps=5)\n",
    "print('\\nTest forecasts')\n",
    "print(future)\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}