notebooks/Part 3 Predictive Modeling.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 3: Predictive Modeling #\n",
    "\n",
    "Ultimate is interested in predicting rider retention. This sample dataset shows the behavior of a cohort of users who signed up for an Ultimate account in January 2014. The data was pulled several months later. A user is considered \"retained\" if they took a trip in the prceeding 30 days.\n",
    "\n",
    "In this notebook I will use this data to undrestand what factors are the best predictors for retention and offer suggestions to operationalize those insights to help Ultimate."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cleaning and EDA ##\n",
    "\n",
    "First I'm going to read in the data and examine it for outliers and missing data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [],
   "source": [
    "# standard imports\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from datetime import datetime, timedelta\n",
    "\n",
    "# plotting imports\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "# machine learning imports\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix, classification_report\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.preprocessing import StandardScaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>city</th>\n",
       "      <th>trips_in_first_30_days</th>\n",
       "      <th>signup_date</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>last_trip_date</th>\n",
       "      <th>phone</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>ultimate_black_user</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>King's Landing</td>\n",
       "      <td>4</td>\n",
       "      <td>2014-01-25</td>\n",
       "      <td>4.7</td>\n",
       "      <td>1.10</td>\n",
       "      <td>2014-06-17</td>\n",
       "      <td>iPhone</td>\n",
       "      <td>15.4</td>\n",
       "      <td>True</td>\n",
       "      <td>46.2</td>\n",
       "      <td>3.67</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Astapor</td>\n",
       "      <td>0</td>\n",
       "      <td>2014-01-29</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2014-05-05</td>\n",
       "      <td>Android</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>50.0</td>\n",
       "      <td>8.26</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Astapor</td>\n",
       "      <td>3</td>\n",
       "      <td>2014-01-06</td>\n",
       "      <td>4.3</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2014-01-07</td>\n",
       "      <td>iPhone</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>100.0</td>\n",
       "      <td>0.77</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>King's Landing</td>\n",
       "      <td>9</td>\n",
       "      <td>2014-01-10</td>\n",
       "      <td>4.6</td>\n",
       "      <td>1.14</td>\n",
       "      <td>2014-06-29</td>\n",
       "      <td>iPhone</td>\n",
       "      <td>20.0</td>\n",
       "      <td>True</td>\n",
       "      <td>80.0</td>\n",
       "      <td>2.36</td>\n",
       "      <td>4.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Winterfell</td>\n",
       "      <td>14</td>\n",
       "      <td>2014-01-27</td>\n",
       "      <td>4.4</td>\n",
       "      <td>1.19</td>\n",
       "      <td>2014-03-15</td>\n",
       "      <td>Android</td>\n",
       "      <td>11.8</td>\n",
       "      <td>False</td>\n",
       "      <td>82.4</td>\n",
       "      <td>3.13</td>\n",
       "      <td>4.9</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             city  trips_in_first_30_days signup_date  avg_rating_of_driver  \\\n",
       "0  King's Landing                       4  2014-01-25                   4.7   \n",
       "1         Astapor                       0  2014-01-29                   5.0   \n",
       "2         Astapor                       3  2014-01-06                   4.3   \n",
       "3  King's Landing                       9  2014-01-10                   4.6   \n",
       "4      Winterfell                      14  2014-01-27                   4.4   \n",
       "\n",
       "   avg_surge last_trip_date    phone  surge_pct  ultimate_black_user  \\\n",
       "0       1.10     2014-06-17   iPhone       15.4                 True   \n",
       "1       1.00     2014-05-05  Android        0.0                False   \n",
       "2       1.00     2014-01-07   iPhone        0.0                False   \n",
       "3       1.14     2014-06-29   iPhone       20.0                 True   \n",
       "4       1.19     2014-03-15  Android       11.8                False   \n",
       "\n",
       "   weekday_pct  avg_dist  avg_rating_by_driver  \n",
       "0         46.2      3.67                   5.0  \n",
       "1         50.0      8.26                   5.0  \n",
       "2        100.0      0.77                   5.0  \n",
       "3         80.0      2.36                   4.9  \n",
       "4         82.4      3.13                   4.9  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#import json file into a pandas dataframe\n",
    "\n",
    "df = pd.read_json('../data/raw/ultimate_data_challenge.json')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 50000 entries, 0 to 49999\n",
      "Data columns (total 12 columns):\n",
      " #   Column                  Non-Null Count  Dtype  \n",
      "---  ------                  --------------  -----  \n",
      " 0   city                    50000 non-null  object \n",
      " 1   trips_in_first_30_days  50000 non-null  int64  \n",
      " 2   signup_date             50000 non-null  object \n",
      " 3   avg_rating_of_driver    41878 non-null  float64\n",
      " 4   avg_surge               50000 non-null  float64\n",
      " 5   last_trip_date          50000 non-null  object \n",
      " 6   phone                   49604 non-null  object \n",
      " 7   surge_pct               50000 non-null  float64\n",
      " 8   ultimate_black_user     50000 non-null  bool   \n",
      " 9   weekday_pct             50000 non-null  float64\n",
      " 10  avg_dist                50000 non-null  float64\n",
      " 11  avg_rating_by_driver    49799 non-null  float64\n",
      "dtypes: bool(1), float64(6), int64(1), object(4)\n",
      "memory usage: 4.2+ MB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It looks like the Average Rating of Driver, Average Rating by Driver, and Phone information are missing for some records. For the average rating, I can interpolate the overall average into the missing rows. Where phone is missing, I will drop those rows, as phone type might be an important feature for predicting retention."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 49604 entries, 0 to 49999\n",
      "Data columns (total 12 columns):\n",
      " #   Column                  Non-Null Count  Dtype  \n",
      "---  ------                  --------------  -----  \n",
      " 0   city                    49604 non-null  object \n",
      " 1   trips_in_first_30_days  49604 non-null  int64  \n",
      " 2   signup_date             49604 non-null  object \n",
      " 3   avg_rating_of_driver    49604 non-null  float64\n",
      " 4   avg_surge               49604 non-null  float64\n",
      " 5   last_trip_date          49604 non-null  object \n",
      " 6   phone                   49604 non-null  object \n",
      " 7   surge_pct               49604 non-null  float64\n",
      " 8   ultimate_black_user     49604 non-null  bool   \n",
      " 9   weekday_pct             49604 non-null  float64\n",
      " 10  avg_dist                49604 non-null  float64\n",
      " 11  avg_rating_by_driver    49604 non-null  float64\n",
      "dtypes: bool(1), float64(6), int64(1), object(4)\n",
      "memory usage: 4.6+ MB\n"
     ]
    }
   ],
   "source": [
    "df['avg_rating_of_driver'] = df['avg_rating_of_driver'].fillna(df['avg_rating_of_driver'].mean())\n",
    "df['avg_rating_by_driver'] = df['avg_rating_by_driver'].fillna(df['avg_rating_by_driver'].mean())\n",
    "df = df.dropna()\n",
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "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>trips_in_first_30_days</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>49604.000000</td>\n",
       "      <td>49604.000000</td>\n",
       "      <td>49604.000000</td>\n",
       "      <td>49604.000000</td>\n",
       "      <td>49604.000000</td>\n",
       "      <td>49604.000000</td>\n",
       "      <td>49604.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>2.277256</td>\n",
       "      <td>4.601482</td>\n",
       "      <td>1.074991</td>\n",
       "      <td>8.875314</td>\n",
       "      <td>60.883832</td>\n",
       "      <td>5.785418</td>\n",
       "      <td>4.777862</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>3.793300</td>\n",
       "      <td>0.565227</td>\n",
       "      <td>0.222666</td>\n",
       "      <td>19.989018</td>\n",
       "      <td>37.086239</td>\n",
       "      <td>5.669709</td>\n",
       "      <td>0.446220</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>4.500000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>33.300000</td>\n",
       "      <td>2.420000</td>\n",
       "      <td>4.700000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>4.700000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>66.700000</td>\n",
       "      <td>3.880000</td>\n",
       "      <td>5.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>3.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>1.050000</td>\n",
       "      <td>8.700000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>6.930000</td>\n",
       "      <td>5.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>125.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>160.960000</td>\n",
       "      <td>5.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       trips_in_first_30_days  avg_rating_of_driver     avg_surge  \\\n",
       "count            49604.000000          49604.000000  49604.000000   \n",
       "mean                 2.277256              4.601482      1.074991   \n",
       "std                  3.793300              0.565227      0.222666   \n",
       "min                  0.000000              1.000000      1.000000   \n",
       "25%                  0.000000              4.500000      1.000000   \n",
       "50%                  1.000000              4.700000      1.000000   \n",
       "75%                  3.000000              5.000000      1.050000   \n",
       "max                125.000000              5.000000      8.000000   \n",
       "\n",
       "          surge_pct   weekday_pct      avg_dist  avg_rating_by_driver  \n",
       "count  49604.000000  49604.000000  49604.000000          49604.000000  \n",
       "mean       8.875314     60.883832      5.785418              4.777862  \n",
       "std       19.989018     37.086239      5.669709              0.446220  \n",
       "min        0.000000      0.000000      0.000000              1.000000  \n",
       "25%        0.000000     33.300000      2.420000              4.700000  \n",
       "50%        0.000000     66.700000      3.880000              5.000000  \n",
       "75%        8.700000    100.000000      6.930000              5.000000  \n",
       "max      100.000000    100.000000    160.960000              5.000000  "
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looking at mean, standard deviation and maximum shows outliers in the trips_in_first_30_days column and avg_distance. It seems very unlikely that someone could take more than two trips per day for 30 days, or that someone's average distance would be more than 100 miles."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Users taking more than 60 trips in the first 30 days: 4\n",
      "Users with an average distance greater than 100 miles: 1\n"
     ]
    }
   ],
   "source": [
    "print('Users taking more than 60 trips in the first 30 days: {}'.\n",
    "      format(df[df['trips_in_first_30_days'] > 60]['city'].count()))\n",
    "print('Users with an average distance greater than 100 miles: {}'.format(df[df['avg_dist'] > 100]['city'].count()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's look at the relationship between number of trips and average distance to see how much these values are outliers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(df['trips_in_first_30_days'], df['avg_dist'])\n",
    "plt.ylabel('Average Trip Distance')\n",
    "plt.xlabel('Number of Trips in First 30 Days')\n",
    "plt.title('Relationship Between Average Trip Distance and Number of Trips')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It does look like there are two wild outliers, one with a very large number of trips and one with a very long trip. Since these are only two records, I will drop them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "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>trips_in_first_30_days</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>49602.000000</td>\n",
       "      <td>49602.000000</td>\n",
       "      <td>49602.000000</td>\n",
       "      <td>49602.000000</td>\n",
       "      <td>49602.000000</td>\n",
       "      <td>49602.000000</td>\n",
       "      <td>49602.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>2.274828</td>\n",
       "      <td>4.601474</td>\n",
       "      <td>1.074991</td>\n",
       "      <td>8.875382</td>\n",
       "      <td>60.882819</td>\n",
       "      <td>5.782366</td>\n",
       "      <td>4.777862</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>3.753126</td>\n",
       "      <td>0.565236</td>\n",
       "      <td>0.222670</td>\n",
       "      <td>19.989366</td>\n",
       "      <td>37.086537</td>\n",
       "      <td>5.626824</td>\n",
       "      <td>0.446229</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>4.500000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>33.300000</td>\n",
       "      <td>2.420000</td>\n",
       "      <td>4.700000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>4.700000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>66.700000</td>\n",
       "      <td>3.880000</td>\n",
       "      <td>5.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>3.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>1.050000</td>\n",
       "      <td>8.700000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>6.930000</td>\n",
       "      <td>5.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>73.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>79.690000</td>\n",
       "      <td>5.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       trips_in_first_30_days  avg_rating_of_driver     avg_surge  \\\n",
       "count            49602.000000          49602.000000  49602.000000   \n",
       "mean                 2.274828              4.601474      1.074991   \n",
       "std                  3.753126              0.565236      0.222670   \n",
       "min                  0.000000              1.000000      1.000000   \n",
       "25%                  0.000000              4.500000      1.000000   \n",
       "50%                  1.000000              4.700000      1.000000   \n",
       "75%                  3.000000              5.000000      1.050000   \n",
       "max                 73.000000              5.000000      8.000000   \n",
       "\n",
       "          surge_pct   weekday_pct      avg_dist  avg_rating_by_driver  \n",
       "count  49602.000000  49602.000000  49602.000000          49602.000000  \n",
       "mean       8.875382     60.882819      5.782366              4.777862  \n",
       "std       19.989366     37.086537      5.626824              0.446229  \n",
       "min        0.000000      0.000000      0.000000              1.000000  \n",
       "25%        0.000000     33.300000      2.420000              4.700000  \n",
       "50%        0.000000     66.700000      3.880000              5.000000  \n",
       "75%        8.700000    100.000000      6.930000              5.000000  \n",
       "max      100.000000    100.000000     79.690000              5.000000  "
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = df[df['trips_in_first_30_days'] < 100]\n",
    "df = df[df['avg_dist'] < 100]\n",
    "df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot again\n",
    "plt.scatter(df['trips_in_first_30_days'], df['avg_dist'])\n",
    "plt.ylabel('Average Trip Distance')\n",
    "plt.xlabel('Number of Trips in First 30 Days')\n",
    "plt.title('Relationship Between Average Trip Distance and Number of Trips')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The plot shows data more clustered without outliers.\n",
    "\n",
    "The original data set had 50000 records and I've retained 49602 or 99.2% of the data.\n",
    "\n",
    "Now I need to create a column showing whether the user was retained, which is whether their last trip date is less than one month ago. Since I don't know when this data was pulled, I'm going to assume it was pulled on the most recent last trip date."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert date columns from string to datetime\n",
    "df['signup_date'] = pd.to_datetime(df['signup_date'])\n",
    "df['last_trip_date'] = pd.to_datetime(df['last_trip_date'])\n",
    "\n",
    "# while I'm at it, I'll convert ultimate_black_user to 1s and 0s and create dummy columns for catetorical data\n",
    "df['ultimate_black_user'] = df['ultimate_black_user'].apply(lambda x: 1 if x else 0)\n",
    "df = pd.get_dummies(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "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>trips_in_first_30_days</th>\n",
       "      <th>signup_date</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>last_trip_date</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>ultimate_black_user</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "      <th>city_Astapor</th>\n",
       "      <th>city_King's Landing</th>\n",
       "      <th>city_Winterfell</th>\n",
       "      <th>phone_Android</th>\n",
       "      <th>phone_iPhone</th>\n",
       "      <th>retained</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>4</td>\n",
       "      <td>2014-01-25</td>\n",
       "      <td>4.7</td>\n",
       "      <td>1.10</td>\n",
       "      <td>2014-06-17</td>\n",
       "      <td>15.4</td>\n",
       "      <td>1</td>\n",
       "      <td>46.2</td>\n",
       "      <td>3.67</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>2014-01-29</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2014-05-05</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>50.0</td>\n",
       "      <td>8.26</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>2014-01-06</td>\n",
       "      <td>4.3</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2014-01-07</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>100.0</td>\n",
       "      <td>0.77</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>9</td>\n",
       "      <td>2014-01-10</td>\n",
       "      <td>4.6</td>\n",
       "      <td>1.14</td>\n",
       "      <td>2014-06-29</td>\n",
       "      <td>20.0</td>\n",
       "      <td>1</td>\n",
       "      <td>80.0</td>\n",
       "      <td>2.36</td>\n",
       "      <td>4.9</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>14</td>\n",
       "      <td>2014-01-27</td>\n",
       "      <td>4.4</td>\n",
       "      <td>1.19</td>\n",
       "      <td>2014-03-15</td>\n",
       "      <td>11.8</td>\n",
       "      <td>0</td>\n",
       "      <td>82.4</td>\n",
       "      <td>3.13</td>\n",
       "      <td>4.9</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   trips_in_first_30_days signup_date  avg_rating_of_driver  avg_surge  \\\n",
       "0                       4  2014-01-25                   4.7       1.10   \n",
       "1                       0  2014-01-29                   5.0       1.00   \n",
       "2                       3  2014-01-06                   4.3       1.00   \n",
       "3                       9  2014-01-10                   4.6       1.14   \n",
       "4                      14  2014-01-27                   4.4       1.19   \n",
       "\n",
       "  last_trip_date  surge_pct  ultimate_black_user  weekday_pct  avg_dist  \\\n",
       "0     2014-06-17       15.4                    1         46.2      3.67   \n",
       "1     2014-05-05        0.0                    0         50.0      8.26   \n",
       "2     2014-01-07        0.0                    0        100.0      0.77   \n",
       "3     2014-06-29       20.0                    1         80.0      2.36   \n",
       "4     2014-03-15       11.8                    0         82.4      3.13   \n",
       "\n",
       "   avg_rating_by_driver  city_Astapor  city_King's Landing  city_Winterfell  \\\n",
       "0                   5.0             0                    1                0   \n",
       "1                   5.0             1                    0                0   \n",
       "2                   5.0             1                    0                0   \n",
       "3                   4.9             0                    1                0   \n",
       "4                   4.9             0                    0                1   \n",
       "\n",
       "   phone_Android  phone_iPhone  retained  \n",
       "0              0             1         1  \n",
       "1              1             0         0  \n",
       "2              0             1         0  \n",
       "3              0             1         1  \n",
       "4              1             0         0  "
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# create column 'retained' that records a 1 if the last trip date was less than one month before the \n",
    "# most recent last trip date\n",
    "\n",
    "most_recent_trip_cutoff = df['last_trip_date'].max() - timedelta(days=30)\n",
    "\n",
    "df['retained'] = df['last_trip_date'].apply(lambda x: 1 if x > most_recent_trip_cutoff else 0)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Percentage of users retained: 36.65%\n"
     ]
    }
   ],
   "source": [
    "# calculate percentage of retained users\n",
    "pct_retained = 100*df[df['retained'] == 1]['retained'].count()/df['retained'].count()\n",
    "print('Percentage of users retained: {}%'.format(round(pct_retained, 2)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I'd like to look at the correlation between retention and other columns."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "c= df.corr()\n",
    "retained_c = pd.DataFrame(c['retained'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAccAAAD8CAYAAAACNm9oAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nOydd3hVVdaH319CICC9VwERKQLSLKgIKCr2LrZxHHUcHcuoo2IXnVHsjh35bGPFEQWxggUQkSLSu1Rp0jsEQrK+P/ZOuOk35BKSy36f5zw5Z5+120Xvurutn8yMQCAQCAQCe0jY3w0IBAKBQKCkEZxjIBAIBALZCM4xEAgEAoFsBOcYCAQCgUA2gnMMBAKBQCAbwTkGAoFAIJCN4BwDgUAgUGKR9Kak1ZJm5PFekl6QNF/SNEkdY1FvcI6BQCAQKMm8DfTK5/1pQHN/XQe8GotKg3MMBAKBQInFzH4E1udjcg7wjjnGAVUl1StqvWWKWkCgdLBh/schFFIgEIiKaodepKLkV9cO0X/f/DTlb7gRXwYDzGxAIaprACyNeF7m01YWoowcBOcYCAQCgf2Gd4SFcYbZyc2RF3kwEJxjIBAIBGJLQrGu2C0DGkU8NwRWFLXQsOYYCAQCgdiSmBj9VXSGAlf6XavHAJvMrEhTqhBGjoFAXDJ24jyeG/AV6enpnH1KJ668uFuW9x8MHsPQYRNJTEygWpWDuO/W86hXuxrzFqzkyVeGsm37ThISxFW9u3PyCW33Uy8CpZYYjhwlfQh0B2pKWgY8BCQBmFl/4CvgdGA+sB34SyzqDc4xEIgz0tLSefrVz3nh33+hds3K/OW2/nQ9phVND66dadPikHq8/Z8bSE4uyydfjuelN4fx6N2XkJycxIO3X8DBDWqyZt1mrvrHKxzT8VAqVSy/H3sUKHUods7RzC4t4L0BN8asQk++PZBUVdLf83n/c6waIukRST33Il85Sd9JmiKpt6TXJbUuRP5zC7KX9C9/uHSKpOGS6vv0vT58KultSRdGax8IRMusectoWL8GDepVJympDCef0JYfx83OYtPpiENITi4LQJuWjVi9djMABzeoycENagJQq0ZlqlWtyIZN24q3A4HST0JC9FcJpaCWVQVyOEdJiQBmdmysGmJmD5rZd3uRtQOQZGbtzewjM7vWzGZlN8pocy6cCxTkTJ8ys3Zm1h74AnjQp++Tw6eBQFFYs24ztWtWyXyuXbMya9ZtztP+8+G/0qVz8xzpM+cuIzU1jYb1qu+TdgbimAPAOT4ONPMjpl8kjZD0ATAdQNJW/7e7pB8lDZY0S1J/SQmSEv0IaYak6ZJuy6uiyJGUpMWSHpY0yedrmUee2sB7QHvfxmaSRkrqnNE+PyIdD3SR9Lhv3zRJT0s6FjgbeCojf271mFnkN8tB7NkmHPXhUz/KfMnX/yVQO+Ldg/7znSFpgLdtJmlShE1zSb/6+yz9yOczvU7SREkT3x64N787AqURy3UTe+7H1r7+YQqzf1vOFRd0zZK+dv0WHn5mEA/cdj4JJfgLLFBCkaK/SigFrTneDbQxs/aSugNf+udFudgehRuBLQG+Ac4HFgENzKwNuGnaQrRtrZl19NO6dwDXZjcws9WSrgXuMLMzfR2RJgcBM8zsQUnVgTeAlmZmkqqa2UZJQ4EvzGxQfo2R9ChwJbAJ6OGTC3P49DygBdAWqAPMAt70714ys0d8Pe8CZ5rZ55I2SWpvZlNwi8xv+36cF9mPvNoceX4oBAE4cKhdszKr127KfF69djO1alTKYTdh8nze/mgUrz5xDWWT9nwVbNuewu193+Fvf+pJm5aNcuQLBAoksfRvZynsT8IJeTjGjHcLzSwN+BA4HlgIHCLpRUm9gLzndnLyqf/7K9CkkO3MIA34xN9vBlKA1yWdj9vVFDVmdp+ZNQLeB27yyYU5fHoC8KGZpZnZCuCHiHc9JI2XNB04ETjcp78O/MVPCfcGPihqPwLxT6vDGrB0+TpW/LGe1NTdfPvjdLoenXXyZe6CFTzx0mc89eDlVK9aMTM9NXU3ff79Aaef2IGTurYp7qYH4oU4mFYtrHvPb2U+u1MwM9sg6QjgVNxuoouBq6Osa6f/m8be76pN8c4aM9st6SjgJOASnIM7cS/K/AA3gn6Iwh8+zeE4JSUDrwCdzWyppL5Asn/9ia/nB+BXM1vn88SiH4E4pUxiInfccCb/eOC/pKenc+bJnTikcR0GvPsdLZs34IRjWvHiG9+wPWUX9/UbCECdWlV5+qEr+G70DCbPWMymzdv58js3q//AbRdwWLMih6oMHEiUYKcXLQU5nS1AzvmY3DlKUlPctGpvYICkmsAuM/tE0gJcdPX9gqSKQAUz+0rSONyZGIiij5Kam9lv/vFsYI6/HwrcJGkgcDT5Hz79EfibpHdw6409cI42wxGu9W28EBgEYGYpkobhNvpcU0A/AoFMjj2yBcce2SJL2nV/2rMZ/KXHcv+NetqJ7TntxPb7tG2B+EcleC0xWvJ1jma2TtIYOR2tHcCqfMzH4jbwtMU5gsH+/i0p89DLPUVv8l5TCfjMj9QEZGwOGgj8n6RbgAvNbEEueR+X1AJIxzn/6316YQ6fDsaN8KYD84BRAH7d8/98+mLgl2z53set3w4voB+BQCBQMoiDkaMs961thSvEbdbJ3BQTiB2S7gCqmNkDRSknbMgJBALRUlRVjoTeZ0f9fZP+0dASOcws/VuK4hhJg4FmxGBNcewb9xe9QYFA4IDg9H4XFa2AOBg5xsQ5mtlIYGQ0tpJeBo7Llvy8mb1VQL6/AP/IljzGzGIWNmhv25atjLbAu9mSd5rZ0YVtj5mdV9g8gUAgsN9JKJGDwUJR7CPHvXVm3kFF7aT2so4iO1ozmw6EHQ2BQODAJYaxVfcXYVo1EAgEArElTKsGAoH9Teuz7qNWi26k7Uph2qC72bwiR2hhKtc/nCMu6kdCUjJr5o5i1uePApBUvgodLn2O8tUasGPDciZ9cCu7UzZTJrky7S58jArVDyZ9906mfXIvW1e500xNjvszjY68CMzYsmoe0wbdQ/ruXcXa50AJJw6cY+nvQSBwAFOrxQlUqNGEUU+fwozBD9Dm3L652rU5ty/TBz/IqKdPoUKNJtQ67AQADul2HWsXjGXUM6eydsFYmnW/DoBDe1zP5pWz+emFs5n6cR9an3kfAOUq16bJsVcy5qULGP38WUiJ1Gt3RrH0NVCKKF6x431CqXWOPtj5sRHP10u6spjqvkjSbEkjorTPDIaey7tCSWwFApHUaXUSyycPAWDj0qmUSa5MuUq1stiUq1SLMuUqsvH3KQAsnzyEOq1Pcvlbn8TySS7/8klDqNPaBQqoWLsZ6+aPA2DbmoWUr9aAshVrAKCERBKTkt3fssns3LJ633c0ULo4AMPHFSuSypjZ7jxedwe2Aj9DpiJ0cXEN8Hczi8o55oWkRDPLEVC9CGWlxaKsQOkhuUodUjb+kfmcsukPkivXYeeWNXtsKtchZXM2myp1AChXsUam7c4tayhX0clTbV45h7ptTmbDkl+p0rAt5avWJ7lyXTavmMmi0W/So88I0lJ3snb+GNb+NqY4uhooTcTBhpy97oGkIZJ+lTTTSyPdIOnJiPdXSXrR3z8gaY6kbyV96A+251XuSEmPSRoF/EPSWT4o92Q5UeM6kprgotTc5qWmukrqm1GuL+MJSRMkzZPU1adXkPQ/L/X0kS831xGdt79UTjJrhqQnfNqDuKDq/SU9lUe+8pIGZtQDlI94l11Ga6SkzgV8flf4vkyR9JoPRJ6jrPz+vQLxSs4t8zkCe+QSyqug4B8LRw0gKbkyx988hCbH/onNK2dj6bspk1yZ2q1PYuRTJ/FDv64kJpWnfvuzi9SDQPyhhISor5JKUUaOV5vZeknlcSHPTgLGAHf5972BR73zuQAnSlwGmIRT2siPqmbWDUBSNeAYL890LXCXmf1TUn9gq5k97e1Oyt43MztK0um44N09ccLNG8ysnaQ2wJS8GiCpPvAE0AnYAAyXdK6ZPSLpRFxEoIl5ZL8B2O7raef7nEGmjJavJyN9EC4EX/bPr5W/P87MUiW9AlwOvJO9rFz6cB1OhJmbetXmtPaFUQwLlFQaH3MZjY68GICNy6aTXLWuC2oIJFepm2Oa040m62Y+J1epy87Nzmbn1nWUq1TLjRor1WLn1vUA7N65jWmf3JuZp/td37NjwzJqNu/KjvXL2LVtAwCrZg6nWuMOrJgydJ/1N1D6iAcN0KL04BZJU4FxOGWKpsBCScdIqoHTLhyDG2V9ZmY7zGwL8HkUZX8Ucd8QGCYn53Qne+ScCiI3yavjcbFUMbMZwLR88h8JjDSzNX5q932c7FQ0nIATYcbMpmWrJ1JGKxMzW0Pun99JOAf9i6Qp/vmQ/MqKKHOAmXU2s87BMcYPS8Z9wE8vnstPL57Lqlnf0aDDuQBUbXQEu1O2ZJlSBTddunvXNqo2OgKABh3OZdXs7wFYPfsHGnR0+Rt0PJdVs1x6meRKKDEJgEZHXsT6RRPZvXMbOzatoOrBR5CQ5OLl1zi0C1tX5xaOOHAgU6ZMmaivkspetczHUu0JdDGz7ZJG4tQlPsLJUs0BBvvR3t6ESoiUxnoReNbMhvp6+0ZZRm6SV4VpS1FDPOQ1b5WSz9pgXp/ff80st6Dt+ZUVOABYM3cUtVt0o9sd35KeuoNpg/aM9o6/eQg/vegc38whfWl3oT/KMe9H1sz9EYAFowbQ4dL/0KjzhezYuJLJH7ggVBVrN+OIi57A0tPZuno+0z5xu1U3LZ3GHzOGcfxNg7H03WxeOZulEz4iEIgkHkaOe+u2q+CmJ7dLagkc49M/Be7DTfL08Wk/Aa9J6ufrOwP4v0LWtdzf/zkifQtQuZDt/gnnfEb4HaJt87EdDzwvJ7u1AbgU56ij4Ufc1OcIP33bLsp8uX1+3+NUOJ4zs9WSqgOVzGxJlGUG4pyZQx/JNT3DMQJsWj6D0c+flcMmdftGJrxxVY70jb9PYdQzp+Za7m/fvchv30X7v0LgQCQenOPe9uAboIykacC/cFOrmNkGYBbQ2Mwm+LRfcLqHU3Ff/hOBTYWoqy/wsaTRwNqI9M+B8zI25ERZ1itALd/uPrjpzlzb4nUZ7wFG+LZPMrPPoqznVaCir+cuYEI0mfL4/GYB9+PWPKcB3wJBeTYQCJRYEqSor5JKTCSrCqxEqmhmWyVVwI2qrjOzSQXl2wftSASSvIhwM9yo7DAzi/vwHl/d0yJIVgUCgag4vd/cInmtanfeHPX3zYanXiyRHrK4VkMH+GnMZNz6WbE7Rk8F3FRnEm5N8YYDwTEGAoFAcRIP06rF4hzN7LLsaYqBPNRetGMLkONcoz8nWC5b8p+8wkaeSDoVd9wjkkUlUWoqIQ4O5QYCgdJBSd6FGi37rQex1GEsKnujtejzDQOGxbg5gUAgUKqJ9chRUi/geSAReN3MHs/2/mDgv0BVb3O3mX1VlDpLv3sPBAKBQIkils7R7xV5GTgZWIY78z3Ub1bM4H7gf2b2ql/C+4o959v3iuAcA4FAIBBTEmM7cjwKmG9mCwEkDQTOwe3sz8DYc7SvCrCiqJUG5xgIlHJanXkvNVucQPquFKZ/cm8eeo6taXthPxKSyrF27o/M/uIxwOk5HnHJs5l6jlM+vI3dKZv35GvQhi43DGTKwNtZNWM4yVXr0+HyF5ASUGISv499LwQBCOQgxtOqDYClEc/LgOxLYX1xx91uxoXV7FnUSsMujRKIHOHfJlAgNQ87gQo1GjP6mV7MGPIQrc/JNcwurc95iJmDH2L0M72oUKMxNQ9zR4Obdvsr6xaMZfSzvVi3YCyHdPvrnkxKoEWvf2ZR3di5ZQ3j+l/Kzy+dz7hXe3NIt7/mkMgKBBISEqK+vHDFxIjrumzF5XbUI/tRkUuBt82sIXA68G5Rv0PDFzD7VGHkFkmzvDrHQJ/WNzKPV/xo4q/ZPrD4JKCRpGvkVEVGSvo/SS/5PLUkfSLpF39l3/UbOECo0/pEVkx2sSk2LZ1KUl56jskV2bjUxdlfMfmzPXqOrfbkj0wHaNzlClbN/JZdW9dlpllaKpaWCkBCYtlcFT8CgcLEVo2MAe2vAdmKW4aL351BQ3JOm14D/A/AzMbijg3WLEofgnN0XG1mnXDHPG7BRfI5P+J9b+CjbAoj55PLsZBs3A10MLN2OImtgmgBvGNmHYBU4AFcaL6TgZYRds8Dz5nZkb49r0dRdiAOKVe5Djs2RWg1bv6DcpVrZ7OpTcqmVRE2qyhX2ek5ls2m51jW6zmWq1ybOof35PfxA3PUmVylLsfdPITufX5g0Y9v5Ah0HggUZuQYBb8AzSU1lVQWuAQXdS2S33GiDHglo2SgSP9hBufo2FcKI9OA9yVdAeQl2hzJEjMb5++PAkaZ2XozSwU+jrDrCbzkVTqGApUlVcpeWOR0xVeTN0ZRfaD0kcvILUfUq2hsstLqjHuY+80zYOk53qVs+oMxL57Lj8+cSv0O51C2Yo1CtDdwIBDL8HFeFekm3LG52bhdqTO9lm2GmOg/gb/67/EPgausiOHfDvgNOftYYeQMnHzV2cADkg7HOcnIHyXJEfeRaiT51ZXg27sjv8r99MQAgG/ubRXCx8UJBx9zGQ07Xwi4gOLlq9Ql46dPcuW6OSWrNq8iuUqdzOfkynUyNR93ZdNz3OX1HCs3aEP7S54BIKlCVWq2OAFLS2O1l7oCN9Lcuno+1Zp0YtWM4fuqu4FSSKzPOfozi19lS3sw4n4WOYPKFIkwcsxfYeRc3EJvxna8n4CzJCVLqohzfrniF4MbmdkIXPDxqkBFYDHQ0dt0xI1Sc2MC0E1SNUllcNOnGQzH/ZLKqKt99N0NlHZ+H/cBP790Pj+/dD6rZ31P/Q7nAFCl0RGk5qHnmLZzG1W8nmP9DuewatYPgNNzzMhfv8M5rJrt0n98+mRGPdWTUU/1ZNWM4cwa+girZ39Pucp1SCjjgkmVSa5MtcYd2bZmUbH0O1B6iPG06n7hgB854hRGrveKF3OJUBiRNAtoHakwIilDYWQJ+SuMJALvSaqCGwU+Z2YbJX0CXOmnRH8B5uWW2cyWS3oMJ521AnemJ6OuW4CXfZvL4IK5R7OmGYgz1swdRc0WJ3DCP4eRluqOcmRw7E2f8vNLbul85mcP0/bCfiSWKceaeaNZO8/pOS4c9TrtL3uWhp0vJGXTCqZ8cFu+9VWs3YyWp92FYQixaPSbbF31277rYKBUEg/h44pFlSOeKE6FkYi6ygCDgTfNbPDelBWmVQOBQLT0emx2kbYhd37l2ai/byb+/fYSueW59Lv34qc4FUb6Surp6xoODNmHdQUCgUBMKMnTpdESnGMhKU6FETPL8wxlIBAIlFRiHD5uvxCcYwwoSQojeVG2bPinDgQCxUNwjoFAIBAIZCM4x0AgEAgEslE2qfS7ltLfg0AgEAiUKMLIMRAIBAKBbATnGNhvSKoKXGZmr+zvtgRKHtUPPY7mvfqghERWTvqUJT+9keW9EpNofd5jVKrfmtTtG5k56E5SNq6gTPkqtL34WSo1aMMfUz5j3leP7aceBEoz8eAcS38P9jP+gP7+oCrw9/1Ud6AkowRanH4fU9//O+NfPofabU6jQq1DspjU73g+u1M2M+6FM1g67l2a9XSRcdJ372LhiJeYP/zp/dHyQJwQD+HjSm7LihlJB0n6UtJUr7HYW9JiSTX9+84+KHmGJuMAScOBdyRVkPQ/r9v4kaTxXt4KSadIGitpkqSPfUzWvNqwWNITkib461CfXkfSYN+2qZKOBR4HmkmaIumpff35BEoPlRu0Zfv630nZsAxL283qGV9Tq0WPLDY1W/Rg5RSn+rNm1rdUO8QJq6en7mDT75NJ372r2NsdiB8SExKivkoqYVp1D72AFWZ2BoCPifpEPvadgOPNbIcXL95gZu0ktQGm+DJqAvcDPc1sm6Q+wO3AI/mUu9nMjpJ0JfAf4EzgBZx81XmSEnEBzO8G2phZCDoeyEK5yrXZuXmPxuPOzauo3LBdnjaWnkZaylaSKlQldXuQNgsUnbJxEFu15Lrt4mc60NOP3LqaWV4BxTMYGiEZdTwwEMDMZuB0HMEpfLQGxvhA438GGhdQ7ocRf7v4+xOBV335aVG0Dciq5/jFr+ujyRKIC3KGqswZQzkam0Bg70hMVNRXSaX0u/cYYWbzJHUCTgf6+SnTSO3F5GxZotFeFPCtmV1amKbkcV9oIvUcf+jbNnzzHSDs3LyKcpXrZj6Xq1yHXV6/MbvNzs2rUEIiickV2b0jqt9cgUCBlOTp0mgp/T2IEZLqA9vN7D3gaZzm4mLc9Clk1VPMzk84YWR8UPK2Pn0ccFzE2mEFSYcV0JTeEX/H+vvvgRt8GYmSKgNbgEpRdS5wQLFlxQwq1GhMctUGKLEMtducxtq5I7PYrJ07knrtnYh6rdYns2HRhP3Q0kC8EtYc44u2wFOS0oFUnDMqD7wh6V6crmJevAL81+srTsZNq24yszWSrgI+lFTO295PHhqOnnKSxuN+uGSMOP+BUwO5BkgDbjCzsZLGSJoBfG1md+5FnwNxiKWnMe+rx2j/p/5IiayYPJhtaxbQtMeNbFkxk7VzR7Jy8qe0Pq8fx9zyJbt3bGLGoLsy83e59RvKlKuIEpOo2fJEprx7HdvXLNyPPQqUNkqy04uWoOcYA/wmmSQzS5HUDDfSO8zMCrXlT9JioLOZrY11G8O0aiAQiJYT+04v0mLgjT98HvX3zcsnnlUiFx7DyDE2VABGSErCrTPeUFjHGAgEAvFCPIwcg3OMAWa2Begcrb2kwUDTbMl9zKxJLNsVCAQC+4PgHAN7hZmdV9x1tr/s/uKuMhAIHKAE5xgIBAKBQDYSE4NzDAQCgUAgC2HkGAgEAoFANuIhfFzp70EgEMjB2F9/4z//9xVp6cbZJ3fkyotOyPL+wyFjGDp8EomJCVStXIH7/nEe9WpXBeDWh95h5txltGt1MM88dMX+aH6glFOS1TaiJTjHQCDOSEtL55n+X/D8v/5M7RqVufr21+h6dEuaHlw70+awQ+rx1rN/Izm5LJ9+NYGX3xrOv/tcDMDl5x9Hys5Uhnw9cX91IVDKSVSJPLpYKPaJe5f0tqQL/f2tkipEvPvKC/UWtY72kk7fy7xXSXopj3db97LMzD4HAvuTWb8to2G96jSoW52kpDL0PKEtP46fk8WmU7tDSE4uC8DhLRqxet2euKpHHtGMg8qXIxDYWxKVEPUVDZJ6SZorab6ku/OwuVjSLEkzJX1Q1D4Ux9j3VtwheQDM7HQzi4UuTntckPADHjlK/zxGICasWbeF2jWrZD7XrlGZNes252n/+be/0qVT8+JoWuAAIVGK+ioIH4HsZeA0nMrRpT6GdaRNc+Ae4DgzOxznd4pEkb5QJTXxsT0znu+Q1Dfi+RagPi56zAiftlhSTZ93jqTXvbjw+5J6+nihv0k6ytsfJelnSZP93xaSyuI0EXt7sd/eXqz4TUm/eNtzCmh+I0nf+F8jD+XSt4qSvvcixdMjy5N0pRc2nirp3Vzy/suPJHP9fJW3iHI3358pvg+VfPqdvl/TJD0c8dnPlvQKMAlolEs9mZJV//3ouwI+jkC8kFtISOXxJfTNiKnMmb+Cy88/fl83K3AAEUvnCBwFzDezhT7y2EAg+/f7X4GXzWwDgJmtpojs0zVHM3tB0u1AjzzihR4KXARcB/wCXIbTRjwbuBc4F5gDnGBmuyX1BB4zswskPYiLQ3oTgKTHgB/M7Go/bTtB0ndmti1ntYD7wNsA24FfJH1pZpGLLCnAeWa22TuycZKG4n653If7hbJWUvXIQiU9CVQB/mKFD1x7B3CjmY2RVBFIkXQK0Ny3V8BQSScAvwMtfD1/z62wSMmq9fM+CrFVDxBq16zM6rV7pklXr9tMzeo5BVwmTFnA2/8bxSv9rqZsUth+EIgdSQmJUdtKug7nAzIY4L+7MmgALI14XgYcna2Yw3xZY4BEoK+ZfVOYNmdnf/8fscjMpgNImgl8b2YmaTrQxNtUwSleNMfpGyblUdYpwNmS7vDPycDBwOw87L81s3W+7k9xTjnSOQp4zDuidNw/UB2c8PCgDGdvZpEqwg8A480s8h+6MIwBnpX0PvCpmS3zzvEUnNoHQEWcs/wdWGJm4/ayrkCc0qp5A5auWM+KPzZQq0YlvvtxOg/fcVEWm7kLVvLky0N57uErqV614n5qaSBeKcw5x8gf8XmQ2/Ay+4/9Mrjvxe5AQ2C0pDZFWcIrqnOMFAOGnILABbEz4j494jmdPW37FzDCzM6T1AQYmUdZAi4ws7lR1p39w83+fDlQC+hkZqleMSPZ15PXKOwXoJOk6tmcZnZyFVE2s8clfYlbSx3nR8oC+pnZa5EF+M8ir1Fx4ACmTGIi/7z+DG596B3S09M5s2dHDmlcmwHvfU+r5g3oenRLXnprGNtTdnHf4x8BUKdWFZ564HIAru/zOkuWrWV7yi7Ovupp7r3lHI7pGNYkA9ET492qy8i6bNQQWJGLzTgzSwUWSZqLc5a/7G2lRXWOq4DakmoAW4EzgexD2QxR3r2VYaoCLPf3V+VSbgbDgJsl3exHnx3MbDJ5c7KfEt2Bm769Opd6V3vH2ANo7NO/BwZLes7M1mVzhN/4dnwp6RQfkDw3FuNElL8mQkRZUjM/kp4uqQvQ0pf3L0nvm9lWSQ1wepOBQJ4c2/kwju2cVVf7uitOyrx/8d9X5Zm3/xPX7qtmBQ4QEmLrHH8BmktqivMFl+CW4CIZgtO/fdsvgx0GFEmEtEgbcryXfgQnBPwFbn0wOwOArzM25OwFTwL9IuaSMxgBtM7YkIMbYSYB0/wmoX8VUO5PwLvAFOCTbOuNAO8DnSVNxI0i5wCY2UzgUWCUpKnAs5GZzOxj4P9wa4Pl86j7YeB5SaNx4sUZ3Oo3J03FOe2vzWw48AEw1k83DyLrj4JAIBAoUcTyKIeZ7QZuwg0UZgP/M7OZkh6RdLY3GwaskzQL5xvuzFg221uC2PEBQtiQEwgEoqX6Yb2LNPR7b/6MqL9vrji0TYmMGIEEx7EAACAASURBVLC/N+QEAoFAIM6Ihwg5ce0cJZ0KPJEteVFx6Skqb1HjYcVRfyQzP3mquKsMBAKllK739C5S/hivOe4X4to5eidU7I4oov5iFzUOBAKB/U2QrAoEAoFAIBthWjUQCAQCgWxEG1C8JBOcYyAQJ1Q7pAuH9LwDJSTwx5QhLBv33yzvlZhEizMfpmK9VqTu2MScIfewc9NKqjY5mibdbyIhMYn0tFQWjXieTUvcySYllKHZKXdRpXEnMGPxqFdYN/eH/dG9QCmiMOHjSirBOQYC8YASaHZKH2YMvJGdm1fR/qp3WP/bj2xftyjTpO4R57A7ZQsT+59HrVan0LT7zcz57F5Sd2xk1qDb2LV1LRVqNqPNJS8y4SUneNPouKtJ3b6BX1+7ABBlylfeTx0MlCbiYVq19I999xGSRkrqXIBNnrqQxYGyaWUGDlwq1T+clA1LSdm4HEvfzZrZw6l+WLcsNjWad2PVjC8AWDPne6o2OQqAbavmsmurC2C1fe0CEsqURYkuhHHddmezdOxbvgRj945NBAIFEWNVjv1CGDmWbm4F3sMpiwQOYMpVrM3Ozasyn3dtWU2l+m2y2JStFGFjaezeuZUy5atkcXg1W5zE1j/mYmmpJJZzAckbn3ADVQ/uxI4Ny1gw/ElSt+cXNjgQiI/dqqW/Bx5Jd8npRyLpOUk/+PuTJL0n6RRJY70+48deEgpJnSSNkvSrpGGS6mUrN0HSfyX92z//RdI8SaOA4yLszpI03uswfiepjs/7m6RaEWXNz9ByzKUPb0vqL2m0r+NMn54o6Wk5Xclpkm5WLlqZuZSXqec4dMKaIn7CgRJNrroFhQuKVKHmITTpcTPzv3nMFZmQSLnKddm8bCqT37qCLcun0/SkImvIBg4AEqSor5JK3DhH4Eegq7/vDFSUlISTopoO3A/0NLOOOGmq2/37F4ELzawT8CYubmoGZXAxVueZ2f3ecT6Mc4on47QdM/gJOMbMOuDEOO8ys3TcyO5yb9MTmJqHtmUGTYBuwBlAf0nJOK2zpkAHM2sHvG9mL+Ai0/cwsx65FWRmA8yss5l1PvuoWvlUGSjt7NyymnKV62Q+l61Um51bs/4g2hVpo0TKlKuYOWosW6k2rS54inmfP0TKRhfnf/eOTaTt2sG6ue6315o531GxToti6E2gtFM2ITHqq6QST87xV5xcVCWc9NVYnJPsigvi3RoYI2kK8GecykYLnODxtz79fpwcSgavATPMLMNhHg2MNLM1XpH6owjbhsAwHxz8TuBwn/4mcKW/vxp4i/z5n5mlm9lvuKjyLXFOtb8PwJtdQzIQYMuKWSRXa0S5KvVRQhlqtTqF9b/9mMVm3W8/UqfNmQDUankSG5c4NZ/EchU5/KL/sHjky2xePjVLnvXzR7udqkDVJkeyfe0iAoGCSEBRXyWVuFlzjNBc/AvwMzAN6AE0AxbhxI0vjcwjqS0w08y65FHsz0APSc+YWUpGVXnYvgg8a2ZDJXUH+vp2LZW0StKJOOd6eR75M7uSy3N+GpKBAFgaC759ijaXvIiUyKppQ9m+diGNu/6NLStns37+j/wx9TNanPUIna8fzO4dm5nz2b0A1O/Um/LVGnHwcddw8HHXADBj4E2kbt/AohEv0OKsRyjT85+kbt/AvC8f3p+9DJQSEkquz4uauHGOnh+BO3AjtOk4OalfgXHAy5IONbP5fodnQ2AuUEtSFzMb66dZD/OyVABvACcAH0s6DyfN9bzXr9wMXARk/NSO1J38c7Z2vY6bXn3XzNLIn4sk/Rc3jXqIb+Nw4HpJI81sd4SGZFG1MgNxxIYFY/h1wZgsaUtG79HItrRdzBlyd458S39+g6U/v5FrmTs3/8G096+LbUMDcY9K8IgwWuJpWhVgNFAPGGtmq4AUYLSZrcEJJX8oaRrOWbb0U6MXAk94DcUpwLGRBZrZs8AknPbjKtyIcCzwnU/PoC/OiY4mp7MaClSk4ClVcM5wFE4I+Xo/Yn0d+B2nVTmVPUKfRdXKDAQCgZgTDxtygp5jMeDPSz5nZl0LsHsb+MLMBsW6DaP7dQ7/0IFAICq63jOxSF5r6rpVUX/fHFGjTon0kPE2rVrikHQ3cAMFrzXuU6ZOm78/qw8EAqWIfH/FR0GZEjwijJbgHPcxZvY48HhkmqT7cOuVkXxsZlcVV7sCgUBgXxEPa47BOe4H/NGQRws0DAQCgVJISV5LjJbgHAOBQCAQU+Jhp2dwjoFAHNH1z0/QuP3J7N61g+9f/TtrFk/NYVOr6RH0vP4VEsuWZ8mUbxn93z4A1Di4DT2ueZak5IPYvGYpw1/+K6k7tpBcsRq9bn2H2s06MGfUB/z49l3F3a1AKSMeRo7x4OADgQDQuP3JVK17CO/d1pER//cPul3zTK523a9+lhGv38p7t3Wkat1DOPiIngCceN0L/DzwYT7scxwLJ35BxzNvAWB36k7Gf/woY95/oNj6EijdxEOEnOAcSxiSmkia4e87S3qhANvL8nofOLBo2ul05oweCMCq+RMpV6EKFarWyWJToWodypavxB+/udBxc0YP5JDOZwBQrd6hrJjtgggsnTaCZkedBcDundtZOXccabt2FldXAqWceJCsCs6xBGNmE83slnxMmrAnIEDgAKdi9XpsXbc883nr+hVUrF4vp836FXts1u2xWbdsNk07OZHjQ485l4o1GhRDqwPxSEIhrpJKSW5biUPSEC9tNdPLQd0g6cmI91dJetHfPyBpjqRvJX0o6Y58yu0kaaqkscCNEendJX3h77tJmuKvyT7A+uNAV592Wy7lZkpWjZm/K4afRKBEksuv8JxBPnKzcX+/f+0m2p5yLRc/OpKk8hVJ3526DxoZOBCIhwg5YUNO4bjazNZLKg/8ApwEjAEydij0Bh71EXEuADrgPuNJuBivefEWcLOZjZL0VB42dwA3mtkYr0WZAtwN3GFmZ+aWwcwG4ELM8dKlVUOEnDik7cnX0vpEF8p39cJJWUZ7FavXZ9uGP7LYu9Fk/T02NeqzbcNKADau+I2h/c4HoGrdZjRpf8q+bn4gTinJa4nREkaOheMWH9t0HNAIFxx8oaRjfDDyFjhneTzwmZntMLMtwOd5FSipClDVzEb5pHfzMB0DPOtFjqtmyFcFDmymf/s6H93TlY/u6crCiV/SsuslANQ5tDO7tm9m+8ZVWey3b1zFrpSt1Dm0MwAtu17Col+/AqB8Za/BLdH5vDuZ8X00oYADgZxI0V/RladekuZ6sfic0fP32F0oyfwApUiEkWOUeBmqnkAXM9suaSSQjNN0vBiYAww2M5MKNVcQlRyVmT0u6UvgdGCcpJ6F7EIgzlkyeTiN25/Mn/4zmd07t/P9a5kz9PTuN5qP7nFBwUa9eTsnXf8KZfxRjiVTvgWg+bEX0u6UawFYMOFzZo98LzP/lS9Mo2z5SiSUSeKQzmfwWb/z2bB8bjH2LlCaKKPYjbskJQIv4wTmlwG/SBpqZrOy2VUCbsGpJxWZ4ByjpwqwwTvGlsAxPv1T4D5gCdDHp/0EvCapH+4zPgP4v9wKNbONkjZJOt7MfiKPGKySmpnZdGC6pC44EeSlOMmqQACAH9+6M9f0DMcIsHrhFD6869gcNtO+6c+0b/rnmv+dW9rFpoGBA4IYT0keBcw3s4UAkgYC5wCzstn9C3gStwRVZMK0avR8A5Txklf/wk2tYmYbcP9Ijc1sgk/7BSdTNRXnPCcCm/Ip+y84vcmxwI48bG6VNMNP6+7ASVpNA3b7zTw5NuQEAoHA/qAw5xwjNw76K7uAaAPcQCCDZT4tE0kdgEZm9kWs+hBGjlFiZjuB0/J4l9uGmKfNrK8XVv4RyP1Etsv/K3BERFJfnz4SGOnvb84j+0kFND0QCASKlcIsLEVuHMyruNyy7alLCcBzOM3emBGc475jgKTWuHXJ/5rZpIIyBAKBQDwQ492qy3AbIDNoCKyIeK4EtAFG+u0edYGhks42s4l7W2lwjvsIM8txOF/Sy8Bx2ZKfN7N9vi2wYf0a+7qKQCAQAGIeW/UXoLmkpsBy4BIigp+Y2SagZsaz3yx5R1EcIwTnWKyY2Y0FWwUCgUDpRpYWs7LMbLekm4BhQCLwppnNlPQIMNHMhsassgiCcwwEAoFAjEmPaWlm9hXwVba0B/Ow7R6LOoNzDAQCgUBssdg6x/1BcI6BQBzR9twHqdOqO2m7djBp4F1sWj4zh02Vhm3oeMmTJCYls2r2SKYPeQSAw8+8m7qHn0j67lS2rfudyQPvIjVlCxWqNeCkPsPZunohAOuXTGHqJ0G+KpAfwTkGAoESQp2W3alYswnf9TuRage354gLHuHHFy7IYdf+gkeY8vF9bFgymS7Xvkntlt1YPWcUq+f9xKyvnsLS02h9xl00P+kGZn3p4upvW/s7I549q7i7FCitxMHIsdQFAfBKFcdGPF8v6coY17E1xuVlqmvk8i5fzcZAIFrqtunJ778OBmDD71NIKl+ZcpVqZbEpV6kWZZIrsmHJZAB+/3Uw9dqcDMCaeT9h6W4jxYYlUyhftW4xtj4QX6QX4iqZlMiRo6Qy+QTW7g5sBX4GMLPc412VAnw/J+Ii6MSirBCM/ACmfJU67Ni45/hXyqY/KF+lLju3rImwqcuOjXuUOlI2rqR8layCyACNj7qQZVO+zHyuUL0h3W8fyu6Urcz++lnWLSryf7KBeCYOvooK7RwlDcEdyEwGnsdtrW1qZnf591cBnczsZkkP4GKFLgXWAr+a2dN5lDsS5/COwx3gnAfcD5QF1vlyygPXA2mSrgBuxkWI2WpmT/syxgM9gKrANWY22kepeRsXj3Q2TiT4xvzOwUh6xpezAXeupjLwsZl19O+bAwPNrFMe+XsB//H9nhSR3heo79uwVtIAXCzAs4GFQHsz2+ht5/vPIx3oDxzsi7nVS1dlKYtswsc+DNN1ADf0rMkp7Srn1d1AXJCLVmP2mPa5xhrJanPYSX8nPT2NZZM+AyBl8xqG/bsrqds3UqVhG47+S39+eLIXu3fGdIIlEE/EwbTq3owc95WmITgppm4AkqoBx3iVi2uBu8zsn5L6452ht8sePq2MmR0l6XTgIZySxt9xQcPbSWoDTCmgHQcBk3x9DwIPmdlNPkB4ezObgouH+nZumSUl4wKNnwjMxyl3RNIJON7Mdni1D8wsXdJnwHnAW5KOBhab2SpJHwDPmdlPkg7Gnfdplb2s7O2IDMs05J/Ngp5jHNL0uCtocnRvADYsnU75qvXJ+N8suUpdUjZllazasfGPLNOlyVXrsWPz6sznRp3Pp27rHozp/6fMtPS0XaRvd2LZm5bNYPvaJVSs1ZSNy6bvq24FSj2l3znuzZpjzDUNI4h0Ig2BYZKmA3cCh0fZvk/9319xIyp8WwYCmNkMXMDu/EiPaMt7Pj/A68BfvIRKb+CDPPK3BBaZ2W/mpNjfy/Z+aG7OzNfZ299fEtGGnsBLkqbgAppX9vIs+ZUVOABYNOY9Rjx7FiOePYuVM4ZzcKfzAKh2cHt2p2zJMqUKsHPLGnanbKPawe0BOLjTefwx4zsAarc4geY9rmPcm38jLTUlM0/Zg6qDlyCqUL0RB9VqwrZ1vxdH9wKlFUuP/iqhFGrkuA81DTPYFnH/IvCsmQ319faNsoyd/m8ae/pX1FhGGaOuT3Cj0R9wU8TrosiTG9vySB8LHCqpFnAu8G+fnoD7zLM4Qf8R51VW4ABj1eyR1GnVnZPv+YHdqSlMHtgn812P2z/P3G069ZMH/VGOcqyaM4pVc0YC0O78viSUKctxf/svsOfIRs1DjqRlr1ux9DQsPY2pgx4gdUd+IjOBQMl1etFS2GnVfaJpmE9dy/39nyPSt+DW/wrDTzjnPcIHA29bgH0CcCFutHmZz4+ZpUgaBrwKXJNP/jlAU6/BuAC4NJpG+h8Vg4FngdkRznc4cBPwFEDE1G4gkIVpn/bNNT3yGMbGZdP54emcAjPf9Tsx17wrpg9jxfRhMWlf4MAgluHj9heFnVbdl5qG2ekLfCxpNG6zSQafA+dJmiKpa645c/IKUMu3uw9uWjW/tmwDDpf0K27d8JGId+/jRoXD88psZim4jTBfSvoJ96MhWj4CriDrFPMtQGdJ0yTNwm1KCgQCgZJJHEyrymzf7dOQVNHMtkZoGl63P6Sb/Bphkh/5NQO+Bw4zs117UdYdQBUzK1UhQsKGnEAgEC3nPrOgSEtRO9bPivr7pnz11jGV8IgV+/qcY0nRNKyAm1JNwq0/3rCXjnEw0Aw3mixVfD9xecFGgUAggNvwUCRK8IgwWvapcywpmoZ+t2znXNoyHiiXLflPZpbrHnUzOy+XMgbjduxG0sfMwiJNIBA4QAnOsdCUJE1DMzs6BmXkcJiBQCBwQBNGjoFAIBAIZONADB8XCARKBxfc8gyHH92LXTu3816/v7Lst5ynf8689mGOOvVyKlSsyh2n1cxMr1a7EVfc+zrlK1YhISGRoa/dz6zxYaUgEC2lf+RY6lQ5AoFAwbQ++lRqNzyURy4/nIFP30jv23MXfpnx85c8/bfjc6SfeuXdTB4xiCevPYa3H/4TF98WhGMChSAOjnIcUM4xUt5K0lWS6hehrA6STNKpUdje6o+zBALFQtvjz2LCsPcBWDxrAuUrVqVy9ZwSVItnTWDz+j9ypJsZyRVcrI3kilXYtG5FDptAIG+CZFWpIpu81VXADGBv/6+/FBc551JcIPD8uBUXX3X7XtZVIEGyKhBJ1Zr12bB6WebzxjXLqVKrfq6OMDe+fuvf3PjMF5xw/g2UK38QL91++r5qaiAeKcEjwmiJa+foR4l34CLaTAMW4LQgF+OOdrwvaQcu9N21GTtPJZ2MOwt5fh7lChde7mRgtKRkH2DgIOB/uKDpibgoQnVwslIjJK01sx6SXgWOxElwDTKzh3y5i3GRcXr4qi4zs/mSGgNvArWANcBfzOx3SW8D63HKJ5OAf2ZrZ6ZkVffmZWhTL3GvPsdA6SPX0MaFCPjRqefFjP/6XX743/M0Ofxo/nTfm/S7qiP7MmhIIH6wAzB8XKlB0uE4p3eimR0B/CPjnZkNwoWzu9zM2gNfAa18wG9wclT5nbs8Dqe6sQAYCWT8rO4FrDCzI8ysDfCNmb2AG532MLMMp3efmXUG2gHdJLWLKHuzmR0FvITTg8Tfv2Nm7XDh6yIXgA4DeppZFsfo+znAzDqbWefgGOOfruf+jT6vj6fP6+PZtG4l1Wo3zHxXtVYDNq1dGXVZXU6/ikkjPgFg8czxJJVN5qAqNQvIFQg40tNSo75KKnHrHHFRbAaZ2VoAM1ufl6GXlXoXuEJSVaAL8HU+ZV+Kl8DyfzMCi08Hekp6QlJXM8srfuvFkiYBk3FSXK0j3n0Y8beLv+/CHnmsd9kjoQVOgLn0/0wLFJnRQ17jiWuP5olrj2ba6KEcderlADRpfRQp2zZFPaUKsGH1Ulp0cr/l6jRuQVLZcmzduKaAXIGAI0PBJZqrpBLP06oif9mo7LyFC2qegnM4ua7f+TitFwBnS7rP11NDUiUzmyepE24k2U/ScDN7JFv+prip3iPNbIOfGk2OMLE87skjPUhWBXIwc9w3tD6mFw9+MIvUndt57/HrMt/1eX08T1zr4l+cc/2jdDqpN0nJFXjk4/mM/fJtvn773wx+uQ+X3vkqPS66GTPjvX7X5VVVIJCDePi9Hs/O8XtgsKTnzGydpOrZ3m8BMgSDMbMVklYA9+PWEvOiJzDVzDJ3qUr6L3CupO+B9Wb2nqStuE0/kXWtxcltbQM2SaoDnIabms2gN/C4/zvWp/2MEz9+F7gcL6EVCOTHx/+5Ndf0DMcI8Fn/+/is/305bP5YMofnbuqRIz0QiIr0sCGnxGJmMyU9CoySlIabwlwcYfI20N9vyMkQEn4fqGVms/Ip+lJgcLa0T4AbgNXAU5LSgVSfBjAA+FrSSr8hZzIwE1gIjMlWVjkf8zWBPdO1twBvSroTvyEnms8gEAgE9gfxMHLcp5JVpQ1JLwGTzeyN/VT/YqBzxjppLLm5W3L4hw4EAlHx4qiUIslIbZj/cdTfN9UOvajAuiT1Ap7HnQJ43cwez/b+duBaYDduAHG1mRVGRzcH8bwhp1B4YeN2uPOIgUAgENhLLD096qsg/D6Pl3FLUK2BS70UYiSTcQOLdsAg4Mmi9iFup1ULi5l1yp5WWEmrGLShyb4oF+DJj0fuq6IDgUAgCzGeVj0KmG9mCwEkDQTOATKXv8xsRIT9OOCKolYanGM+xELSKhAIBA40CnNEIzJYiWeAmQ2IeG4ALI14Xgbk9918DfkfxYuK4BwDgUAgEFMKM3L0jnBAPia5rUnmuqYp6Qpc9LNuUTcgD4JzDAQCgUBMiWYtsRAsAxpFPDckl5jYknrioqJ1M7OdRa00OMdAIA4ZM34aTz7/Punp6Zx3ZjeuvuLMLO/fHfgNg78YRWJiAtWqVqbvPddQv25Nfpk0m6de/CDTbvHvK3n8oRs48YQcS/KBQJ5Y2q5YFvcL0NwHUFmOO/N9WaSBpA7Aa0AvM1sdi0qDcwwE4oy0tHT6PfsO/Z+7izq1qnP5X/vS7bgONGvaINOm5WGNef/1vpRPLsf/Bn/Pf179iCcfvpEjO7bif2/9C4BNm7dy1iV30eWoNvurK4FSSiw35JjZbkk34dSPEoE3/Tn2R4CJZjYUeAqoCHzsg+7/bmZnF6XefX6UI1YaipIWS6rp7ztJWuQ1Fc+WdHcR2jdSUpMobftKumNv68qn3K3+b31Jg2JdfuDAYsbshTRqUIeG9WuTlFSGU086mpE/Tcpic2THVpRPdhux2x1+KKtW5ww9/O3IXzjumHaZdoFAtMTyKAeAmX1lZoeZWTMze9SnPegdI2bW08zqmFl7fxXJMUIxjBxjrKGIV7AYBPQ2s8m48y1Di9LGkoKZrcBJYQUCe83qNRuoW3tPtMQ6taozffaCPO0HfzmK449plyN92Pfj+dPFvfZJGwPxTTxEyInpyFHSlZKmSZoq6V2f1lfSHZIuZI+G4hRJZ0gaHJH3ZEmfFlBFK2AI7qzhBJ/vKh/ZBklvS3pB0s+SFvo6kZQg6RVJMyV9IemrjHc4PcQ0SYk+/wxJ0yXdVoh+D5H0qy//uoj0rZIe9Z/HOB9LFUlNJY2V9Iukf0XYN5E0I6Jfn0r6RtJvkp6MsLtG0jw/6v2/jP7n0q7rJE2UNPGNd4ZE251AKcdy2cinXDf8wZfDxjBrzmL+fGlWMeM1azcyf8EyuhwdplQDhSeockQQoZ94nJmtzR7o28wG+XnjO8xsotzE8DOSaplZRrzQ/DQUAT4DrjCz/AJv18NJOrXEjSgHAecDTYC2QG1gNk48mAxBY6+m0cDrMCInXRUtV5vZeknlgV8kfWJm64CDgHFmdp93bn8F/o0Lg/Sqmb0j6cZ8ym2PEzLeCcyV9CKQBjwAdMQFNP8BmJpb5sgt0jtWjwvh4w4Q6tSqzh8R06Sr1qynVs2c/zmPmziT19/9nDdevJeyZZOyvBs+YgI9TuhIUpmwLSFQeEqyTmO0xHLkGLV+on9fWA1FgO+Aa304obwYYmbpPnh4HZ92PE6GKt3M/gBG5JJvIXCIpBd9HL/NBbQlklskTcVFZmgENPfpu4Av/P2vOAcNTiw5Q7fx3XzK/d7MNplZCi4aRGNctIhRZrbezFKBjwvRzsABwOEtm/L7slUsX7GG1NTdDPt+PN2O75DFZs68Jfz7qbf4T79bqV6tco4yvvluHKf17JIjPRCIBrO0qK+SSix/FhZWPxGi1FCM4CagP/AK8Lc8bCLPtyjb3zzx2opHAKcCNwIXA1cXlE9Sd5yMVRcz2y5pJHv0GVNtT2T3NLJ+3tF8VpF9ychfpIDAgfinTJlE7r7tT9zwz6dIT0/nnDNO4NCmDXnl9U9p3bIJ3Y/vyHOvDGT7jp3c+eDLANSrU53nH3crCctXruGP1evo1L7F/uxGoBRTkqdLoyWWzjGHfmIuo8e91VDMIB0n4zRM0iNm9mCUbfsJ+LOc7mItoDvwQaSB3wm7y8w+kbQAJ2kVDVWADd4xtgSOiSLPGNxZnfdw+oyFYQLwnKRquM/zAmCfxHoNlF66djmCrl2OyJL292vPz7x/7T998szboF4tvh38/D5rWyD+KckjwmiJmXPMQz/xqmxmb7N3GoqR9eyUdI6vZxVOOLggPgFOwu2UnQeMBzZls2kAvCUpY6r5njzKul9SpIpsM+B6SdOAubip1YL4B/CBpH/4tkWNmS2X9BiuDytw063Z+xIIBAL7jXgYOe53PUcVk4aipIpmtlVSDdzo6zi//ljqiOhLGZzw8ptmll2AOQthQ04gEIiW8rWPKdLyzcIR90T9fXNIj34lcqlov25Fk9NQ3Ab8sxiq+8Jv/CkL/Ku0OkZPX7k4gsnAcNzxlnwZ3DccnwwEAtFx2SvLipQ/PTWm4eP2C/vVORanhqKZdS9K/pKEmcU8Sk8gEAjEivQ4mFYtcYeYgoZiIBAIlG4sLTjHQCAQCASyEJxjIBAoUXS66BHqH34iu1N3MO6d29iwdEYOm2qN2tLlyudITEpmxcwf+PXjrCeiWvb8Gx3Pf4BP7mzLzm0baNXzepoceR4ASkykct3mfHrXEezavrFY+hQofVh6QUfWSz7BOQYCcUL9w0+kUu2mfN73eGo06ciRl/Rj+FNn5bA76tJ+TPjgLtYumkT3G9+lXuserJzlgkZVqFaPei27sm3dng0Zs7/rz+zvnH5Ag7Y9aXHiX4NjDORLehyMHPe5ZNW+RjGQxJL0D0n/iXh+TdJ3Ec83S3rB3/8cRXm3SqqwF+1o6YOyT5bULB+7SPmurYWtJxCfNGh3CovGO8WzdYsnUbZCZZIr185ik1y5NmWSK7J2kZOwWjR+EA2PODXzfccL+jJ58KO5Bi8HCPvZsAAAFPhJREFUaNz5XJZM/Gwf9SAQL6Sn7or6KqmUeudoZv3N7B3/eBWwN3qRPwPHRjy3B6pExHA9FhfVBjM7loK5FSiUc/R1nQt8ZmYdzCxvjaFAIBcqVK3L9g171OC2b1hJhap1c9js2LgyV5sGbU9mx6Y/2Lh8dq7lJyYlU691d5ZO/moftD4QT8SDKkepc47KJoul2EhiTQYOk1ReUhVgOzAFp+IBzjn+7MvJECbu7iWjBkmaI+l9OW7BOegRkkZ421O8RNUkSR9LqujTF0t6UNJPQG+cU702It8Vkib4/rxWQMD13D6rTMmqH2ZFE0goUKpRzrPUOYJ85GIDRmJSMof3uoVpnz+dZ/EN2p3M2oW/hCnVQIGkp6VFfZVUStWaYx6yWLdA0SSxzGy3pCnAkUB5XGi234BjJa3GRRJamkvWDsDhuDBuY3y7XpB0O9DDt7EmLnZsTzPbJqkPcDvwiC8jxcyO9/07DNhqZk9LaoVzmMeZWaqkV3BxWN8hSiIlqz74e8MQIScOaX7Cnzn0uMsAWLdkKhWq7Zk4qVCtHjs2rcpiv33DSspXrZfFZvvGVVSs1YSKNRtx2n3DXXrVevS65xuGPXkmKZvXANC40zks/iVMqQYKpiSPCKOlVDlHcpHFUq6/hJ0klpzg8hWS3sJJYl2ZT9ljcCPE8sBYnHO8F1iDHzXmwgQzWwbgnWsTXJDzSI4BWgNjfFvL+vIz+Oj/27vzKCmq843j30eWIIoggyCEKBoUxQ0QUAIuRFRIjOSIUTBRNC5gEow/o8REEyUxRyPHuEQUiQYxKrgSkUUkIqgICoKgiBECisi+CCrbLO/vj7o99DTdM90zw0z3zPs5p89UV92qujVw+s69VX2fFMc+CziZKB+SUK/1pdTf1UJL3xjD0jfGANDq+O9z9BlX8Nm8l8hr04n8HV+xc1vJ/zI7t62nYNfX5LXpxKZP53PEKRfyyYzRbF39MS/+tkNxufP/PJupd/2AXd9sAaBeg0Y0P+pU3n58SNVdnMtZ/lWOqpdpLFYmkVhvE8VgNQBGEDWK7cPPWSn2SRYplazO08xsQIpjpBrvFDDGzFJNgO5cCas/nE6r477Pj4a9ReHuncz51w3F2/r8bipT7owevJk79vecetnfqFOvAWsWz2D14ullHrt1h96sXTKTwt079ln9Xc1RmL+r7EJZLtcax71isRK2VyQS622ixvQLM1sPIGkD0Bf4SYb1jNVjI1FKxwhJbc1sWXiKtbWZfVLGMV4DXgrXuj5cayMz+yzDurhaZN4ztyZdH2sYATavXMTkO3qVepwJfygZdLxiznOsmOO52i493nOsYilisT6NK/I45YzECmHHG4DFcatnA92BhRlWdRQwRdIaM+sp6XJgrKTYnLG3EkVnlVafjyTdCrwaYrTyiUKYvXF0zmW1mnDPsdojq/a1qorEynb+QI5zLl2XPLSqQjFSc+7vmfbnzam/ft0jq6paFUdiOeecw1M5sl5VRmJlu/Gz/UFX51x6Lqng/n7PMQd5JJZzzu1bRTXgadWcmyHHOedcdisqKkz7lQ5JvSX9V9IySTcn2f4tSc+E7e9IalPRa/DG0TnnXKWywsK0X2UJ02aOAPoQffd8gKT2CcWuBLaYWVvgXuCvFb2GWjes6lxtccXQv9Gpe2927dzBiNuuZMXH75fYXr/B/vzm7rG0aP1diooKee+NSTz1wC0AnH3h1fS+6FqKigrZuf1rHrnjF6xannxCcucSFRVV6sPxXYFlZrYcQNI4ou+fx389ry9we1h+HnhQkqwCX8fwnqNzNVDHHr1peVhbhvRtzyN3XMvVv38wabkJT9zL9RecwND+XWh3Ujc6dI8mC3hryjh+c1EnburfhZfG3MPAG+6uyuq7HFdUVJT2Kw3fBuLntl4V1iUtE2ZC2wrkVeQacrZxjM80rIZzd5Rkks4tu/Re+7aRtHc8e+n7tJL0fIptMyR1zrQermbrcsaPmDnxKQCWfvAuBzRqQpNmJeOrdu/cweJ5MwEoKMhnxccLyGsefebs+Oar4nLf2v+AlPmOziVTVGRpv+LTg8LrmoTDJY+SybxMRnxYtXwGEE0wPgCYWhkHlFQ31dyvZrYauLAyzuNqh6bNW7Fp7Z4/tjetW0XT5q34cuPapOUbHtiYk0//IZOe3tPDPPeiwZz3s19Tt159hg3K+O9AV4sVZPBVjvj0oBRWAd+Je9+aKAkpWZlVkuoCjYHNaVciiazvOYae1seSxoQcx+fD/KQAQ0JG4geSjgnlm0r6dyg7R9KJYf3tkv4ZelrLQ+5i7Bxp5yaGKKwLiYKVz5HUIK6eSyT9Q9JiSa9K2j9sO1lR/uRsoingYse6POQ7vkw0TZwkDZf0Ybimi+OO/WFY3l/SuHB9zxCldaSqa/FfZMs3pjV84WqIpGk1KW6/7FenDtff9S8mjx3B+i9WFK+f+uxIhpx/LE/dfwv9rvL57136Muk5pmEucJSkIyTVB/oDExLKTAAGhuULgekVud8IOdA4Bu2AUWZ2IrAN+EVYv9HMOgEPAzeGdcOIpos7kShyKj7/8BjgXKIbvLdJqpeQm9iBKF3jp6XUpTuwwsz+B8wAfhC37ShghJkdB3wJ9AvrRwPXmVnJ2Zwj3YCBZvZ94AKgA3AS0AsYLqllQvlrge3h+v5CFGuVlJmNMrPOZtb5yGa58k/tyuvciwYzfNxcho+by+YNa8g7dM8f23ktWrN5w5qk+w269WHWrFzG5Kf/nnT7rKnP0PXM8/dJnV3NVJn3HMOI2q+IRumWAM+Gebb/JCn2H/MxIE/SMqK83L2+7pGpXBlW/dzMYrFRTxICjoEXw8/3iBoWgB6ERsnMpkvKk9Q4bJtkZruAXYpCjFuQeW7iAGBcWB4HXBpXjxVmFnsk8D2gTTh3EzObGdb/i+iR5JhpZhbr/vcAxppZIbBO0kyiAOZFceVPBx4I17dIUvw2V4tNfXYkU58dCUCnHn3o3f9aZr3yDEed0JXtX29NOqTa/xfDaNioMSP/NKjE+kMPa8valcuiY532A9Z8vmzfX4CrMSr5aVXMbDIwOWHdH+OWd5J5elKpcqVxTPxNx97HpmGIz1Is7cZssvzFtHMTw3BrP+B8SbeEffMkxWKyEo+/P2VnUMbnOaY7Aa8/HeFKNf+tKXTs0Zu/T1jC7p07GHH7VcXbho+by039u9C0+bfpd/XvWLX8Y+4e+y4AU555iOnjR9Pn4ms54ZSzKCzI5+ttW3jwD1dW16W4HFTZjWN1yJXG8TBJ3cxsNnsehumYouwbRMOif5Z0JtHQ67ak92AimeQm9gIWmlnx0wmSxgA/Bt5MdnAz+1LSVkk9zOwtSh+yfQMYFI7ZlKiXeBNRAHPi9b0u6XjgxFKO52qxx+76ddL1N/XvAsDm9V/wk471k5YZPdzn6nflV1CQ+3Or5sqNqCXAwDCE2JToHmMqtwOdQ9m72HOTNqmQ8xjLTVwETAMS7/PFDADGJ6x7gbLn6b2CKPB4NlBalPp4oiHUhcB0YKiZJY6FPQwcGOo6FHi3jHM751yVquQHcqpF1uc5hjnyJprZ8dVclZz2k471s/sf2jmXNZ5bsLtCGYuZ5MdWNDtyX8mVYVVXQY+Nf7HsQs45VwmyuUeYrqxvHM3sU6DKe421NffROecqKs1p4bJa1jeO1cVzH51zrny85+icc84lqAlPq3rj6FwN9Pbcj7ln5L8pKiyib59TuPzis0psf+qFmbz0yjvUqbMfTRofwB9vuJiWLZqyZt1mhv5pDIVFRRQUFHJx3x70O+971XQVLld5z9E5l3UKC4u4e8SLPHjnIFo0a8zAIfdx+qnHceThe1I52n332zzx9+tp0KA+z7/8Ng88OpE7b7mMZk0P4rF7h1C/fl2279hF/0HDOb3bcRyS17iUMzpXUk1oHHPle45pq64oK0mTJTUJy4VhEvMPw8TiDcsTVeVceSz+70q+0yqP1i3zqFevLmef2ZGZsxeXKNO5Q1saNIgmADjh2MNYv3ErAPXq1aV+/ehv5t35BTXiQ85VvUrOc6wW3nOsJGYWPwH5jjCJOZKeAgazZ/5V5/apDZu20uKQJsXvWzRrzIcfr0xZ/qVX3uV7XY4pfr92/Rb+74+P8fnqjVx31Xnea3QZK8ry78+nI2d7jlkYZZWqx/om0DYs10kRadUh1GmRpPGSDg7rZ0j6a6jDJ5JOC+vrhGiruWGfQUnO62qpZJ9LqWZPnPzaeyxZ+jmXXtizeN2hzQ9m7MgbGT/6d0yaNo9NW75KvrNzKdSEnmPONo5BNkVZ7UVR6GYfIPbdyFSRVk8Avw11+wC4Le4wdc2sK3B93Porga1m1oUoteNqSUckOX9xnuPop1/JpOouhzVv1ph1G74sfr9u41aaJen9vTP/E0aP/Q/3DPt58VBqvEPyGnPk4Yfy/ofL92l9Xc1TUFCU9itb5XrjmBhl1SMsx0dZtQnLPYjiojCz6URpGiWirMxsI1FcVWKU1fvh/ZFp1mv/sM88YCVR1hikF2k1hmjC8Zhk13IOcFk4xztAHlHDW0J8nuMVl/ROs+ou17Vv9x1WfrGRL9ZuIj+/gGkzFnD6qceVKPPfZau484HnuWfYz2napFHx+nUbvmTnrnwAtn21nUUfreDw1s2rtP4u99WEuVVz/Z5jVkRZJVF8zzEmpIIki7QqS6prGWJmU8tRN1fD1a1Th6G/vIDrfj+KwiLj/HO68t02hzJyzCsce3Rrzuh2PPf/YyI7duzi5juiAZRDmzfhb8Ou5NOV67jvHy8X56z99MIzaXtEqnn4nUsui9u8tOV645gtUVblZmZbJW2RdJqZvUkUnjyzjN2mAtdKmm5m+ZKOBr4ws2/K2M/VEt27Hkv3rseWWDd44J7Rg4f+Ojjpfqec3I6xJ7fbp3VzNZ83jtUvFmX1CLCU6B7jkBRlbwdGh6in7aQRZSUpFmW1H5AP/BKo1MYxGAiMDA8ULSeKuCrNo0RDrPMVte4biDIlnXOu2hXWgNYx6yOrUvEoq8xs+3Ribv5DO+eq3EFtzqtQjNSwHzZM+/PmtknbPbLKOedczVcDOo652zh6lFVmGl/6h+qugnMuR9ib51Vo/yz++mLacrZxrC4eZeWcc6UrzNHbdfG8cXTOOVepfFjVOeecS1AThlVzfYYc51w5PHbzbayb8BofjHmuuqviaqCCovRf2cobx0om6fq4CdBLK/eopPaVdM5qielyuevxKS/T+8ZfVnc1XA1VaJb2K1t541gOiqT63V0PlNk4mtlVZvZR5dbMufS8uXA+m7dtre5quBqqqCj9V7byxjFNISJriaSHgPnApZJmh2is5yQdGOKuWgGvS3o97PdwSMZYLGlY3PFmSOoclr+W9BdJC0N0VYuw/hBJL4RoqrmSuof1eSHyakGYHSgrv0TrnKudiiz9V0WEKMJpkpaGnwcnKdMhfFYvDhF/F6dzbG8cM9OOKF7qbKLYqF4hGmsecIOZPQCsBnqaWSwg7xYz6wycCJwRy5FMcAAwx8xOIpoD9uqw/n7g3hBN1Y9o2jiIoqveMrOOwATgsGSVjY+sYu3GCl24c86lq6oaR+Bm4DUzO4poPuybk5TZDlwWogJ7A/dJapKkXAn+tGpmPjOzOZLOA9oDs8LE5fWB2Sn2uUjSNUS/65Zhv0UJZXYDE8Pye0SNL0AvoH3c5OgHSWpEFGl1AYCZTZK0JdmJzWwUMApAp3XM3sF951yNUoXDpX2BM8PyGGAG8Nv4Amb2SdzyaknrgUOIMnVT8sYxM7HUCwHTzGxAaYVDAPGNQBcz2yLpcaBBkqL5tmeS2/hoqv2Abma2I+G4sHdcl3POZYX8DLqEofNwTdyqUeEP+3S0MLM1AGa2RlKp4aOSuhJ1Zv5X1oF9WLV85gDdJbUFkNQwxEYBfAXE0mMPImpQt4b7iH0yPM+rwK9ibyTFMiJj8VtI6gPsNc7uXGmevu1OZo8cQ7vDDufzF17h5z/0UBdXeTJ5ICc+lD28SjSMkv4j6cMkr76Z1ElSS6LA+yvMrMy+rfccy8HMNki6HBgrKTbP6q3AJ0TDmFMkrTGznpIWAIuJoqhmZXiq64ARIWarLlGjOBgYFs49nyj7cWVFr8nVLpcMK0+Gt3PpqcwZcsysV6ptktZJahl6jS2B9SnKHQRMAm41sznpnDdnI6tcZvyeo3MuXfbmggo9Af+zLt9K+/Pmybm7yn0uScOBTWZ2l6SbgaZmNjShTH1gCvCymd2X7rF9WNU551ylqsJJAO4Czpa0lOhBxrsAJHWWFHu6/yKihxgvl/R+eHVIfrg9fFjVOedcpSoorJrzmNkm4Kwk6+cBV4XlJ4EnMz22D6s6V4tJuiaDJwOdqzV8WNW52u2asos4V/t44+icc84l8MbROeecS+CNo3O1m99vdC4JfyDHOeecS+A9R+eccy6BN47OOedcAm8cnathJF0vqWEa5R6V1L6SzvmppGaVcSznsoHfc3QuBynKLVOydAFJnwKdzazKEq6r45zO7Uvec3QuR0hqI2mJpIeA+cClkmZLmi/pOUkHSroOaAW8Lun1sN/DkuZJWixpWNzxZkjqHJa/lvQXSQslzQkRa0g6RNILkuaGV/ewPk/Sq5IWSHqEKOPUuRrDG0fncks74AmiSZavBHqZWSdgHnCDmT0ArAZ6mlnPsM8tZtYZOBE4Q9KJSY57ADDHzE4iika7Oqy/H7jXzLoA/YDYZM63AW+ZWUdgAnBYJV+nc9XKJx53Lrd8ZmZzJJ0HtAdmRSOs1Admp9jnopC2XhdoGfZblFBmNzAxLL9H1PgC9ALah3MAHCSpEVHKwQUAZjZJ0paKXphz2cQbR+dyyzfhp4BpZjagtMKSjgBuBLqY2RZJjwMNkhTNtz0PIBSy57NhP6Cbme1IOC6AP7DgaiwfVnUuN80BuktqCyCpoaSjw7avgEZh+SCiBnVruI/YJ8PzvAr8KvYmLgfvDeCnYV0f4ODyXIRz2cobR+dykJltAC4HxkpaRNRYHhM2jwKmSHrdzBYCC4DFwD+BWRme6jqgs6RFkj4CBof1w4DTJc0HzgFWVuR6nMs2/lUO55xzLoH3HJ1zzrkE3jg655xzCbxxdM455xJ44+icc84l8MbROeecS+CNo3POOZfAG0fnnHMuwf8DCEJRPyKQ00UAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.heatmap(retained_c,cmap='BrBG',annot=True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Phone type, city, and number of trips in first 30 days seem to have the strongest correlations with retention."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predictive Model ##\n",
    "\n",
    "Goal is to create a predictive model to help Ultimate determine whether or not a user will be active in their sixth month in the system. \n",
    "\n",
    "First I need to create smaller dataset of users who have been in the system for at least six months, and whether they were active in their sixth month."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "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>trips_in_first_30_days</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>ultimate_black_user</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "      <th>city_Astapor</th>\n",
       "      <th>city_King's Landing</th>\n",
       "      <th>city_Winterfell</th>\n",
       "      <th>phone_Android</th>\n",
       "      <th>phone_iPhone</th>\n",
       "      <th>retained</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "      <td>2723.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>1.654793</td>\n",
       "      <td>4.590373</td>\n",
       "      <td>1.080602</td>\n",
       "      <td>8.643739</td>\n",
       "      <td>0.375689</td>\n",
       "      <td>64.886045</td>\n",
       "      <td>5.645402</td>\n",
       "      <td>4.770814</td>\n",
       "      <td>0.325009</td>\n",
       "      <td>0.183988</td>\n",
       "      <td>0.491003</td>\n",
       "      <td>0.288652</td>\n",
       "      <td>0.711348</td>\n",
       "      <td>0.379361</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>3.145358</td>\n",
       "      <td>0.564544</td>\n",
       "      <td>0.287310</td>\n",
       "      <td>20.354421</td>\n",
       "      <td>0.484389</td>\n",
       "      <td>35.809550</td>\n",
       "      <td>5.259156</td>\n",
       "      <td>0.456343</td>\n",
       "      <td>0.468464</td>\n",
       "      <td>0.387546</td>\n",
       "      <td>0.500011</td>\n",
       "      <td>0.453219</td>\n",
       "      <td>0.453219</td>\n",
       "      <td>0.485317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>4.500000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>42.900000</td>\n",
       "      <td>2.450000</td>\n",
       "      <td>4.700000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>4.700000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>69.900000</td>\n",
       "      <td>3.780000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>2.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>1.040000</td>\n",
       "      <td>7.100000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>6.645000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>54.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>41.830000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       trips_in_first_30_days  avg_rating_of_driver    avg_surge    surge_pct  \\\n",
       "count             2723.000000           2723.000000  2723.000000  2723.000000   \n",
       "mean                 1.654793              4.590373     1.080602     8.643739   \n",
       "std                  3.145358              0.564544     0.287310    20.354421   \n",
       "min                  0.000000              1.000000     1.000000     0.000000   \n",
       "25%                  0.000000              4.500000     1.000000     0.000000   \n",
       "50%                  1.000000              4.700000     1.000000     0.000000   \n",
       "75%                  2.000000              5.000000     1.040000     7.100000   \n",
       "max                 54.000000              5.000000     8.000000   100.000000   \n",
       "\n",
       "       ultimate_black_user  weekday_pct     avg_dist  avg_rating_by_driver  \\\n",
       "count          2723.000000  2723.000000  2723.000000           2723.000000   \n",
       "mean              0.375689    64.886045     5.645402              4.770814   \n",
       "std               0.484389    35.809550     5.259156              0.456343   \n",
       "min               0.000000     0.000000     0.000000              1.000000   \n",
       "25%               0.000000    42.900000     2.450000              4.700000   \n",
       "50%               0.000000    69.900000     3.780000              5.000000   \n",
       "75%               1.000000   100.000000     6.645000              5.000000   \n",
       "max               1.000000   100.000000    41.830000              5.000000   \n",
       "\n",
       "       city_Astapor  city_King's Landing  city_Winterfell  phone_Android  \\\n",
       "count   2723.000000          2723.000000      2723.000000    2723.000000   \n",
       "mean       0.325009             0.183988         0.491003       0.288652   \n",
       "std        0.468464             0.387546         0.500011       0.453219   \n",
       "min        0.000000             0.000000         0.000000       0.000000   \n",
       "25%        0.000000             0.000000         0.000000       0.000000   \n",
       "50%        0.000000             0.000000         0.000000       0.000000   \n",
       "75%        1.000000             0.000000         1.000000       1.000000   \n",
       "max        1.000000             1.000000         1.000000       1.000000   \n",
       "\n",
       "       phone_iPhone     retained  \n",
       "count   2723.000000  2723.000000  \n",
       "mean       0.711348     0.379361  \n",
       "std        0.453219     0.485317  \n",
       "min        0.000000     0.000000  \n",
       "25%        0.000000     0.000000  \n",
       "50%        1.000000     0.000000  \n",
       "75%        1.000000     1.000000  \n",
       "max        1.000000     1.000000  "
      ]
     },
     "execution_count": 103,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "last_trip = df['last_trip_date'].max()\n",
    "date_6mo = last_trip - timedelta(days=180)\n",
    "\n",
    "df_6mo = df[df['signup_date'] <= date_6mo].copy()\n",
    "df_6mo.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This leaves is only 2723 users out of the full data set are older than 6 months. \n",
    "\n",
    "This is a classification problem, i.e. determining whether users are retained or not, so I will start with LogisticRegression and see how that performs.\n",
    "\n",
    "In order to perform machine learning on this data, I need to convert dates to meaningful numbers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "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>trips_in_first_30_days</th>\n",
       "      <th>signup_date</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>last_trip_date</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>ultimate_black_user</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "      <th>city_Astapor</th>\n",
       "      <th>city_King's Landing</th>\n",
       "      <th>city_Winterfell</th>\n",
       "      <th>phone_Android</th>\n",
       "      <th>phone_iPhone</th>\n",
       "      <th>retained</th>\n",
       "      <th>signup_days</th>\n",
       "      <th>last_trip_days</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0</td>\n",
       "      <td>2014-01-02</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>1.17</td>\n",
       "      <td>2014-06-03</td>\n",
       "      <td>33.3</td>\n",
       "      <td>0</td>\n",
       "      <td>66.7</td>\n",
       "      <td>12.08</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>180</td>\n",
       "      <td>28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>1</td>\n",
       "      <td>2014-01-02</td>\n",
       "      <td>4.601559</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2014-01-03</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>100.0</td>\n",
       "      <td>14.48</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>180</td>\n",
       "      <td>179</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>0</td>\n",
       "      <td>2014-01-02</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2014-04-26</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>50.0</td>\n",
       "      <td>1.38</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>180</td>\n",
       "      <td>66</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51</th>\n",
       "      <td>4</td>\n",
       "      <td>2014-01-02</td>\n",
       "      <td>4.500000</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2014-05-02</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>52.4</td>\n",
       "      <td>3.24</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>180</td>\n",
       "      <td>60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>68</th>\n",
       "      <td>0</td>\n",
       "      <td>2014-01-01</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2014-06-15</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4.44</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>181</td>\n",
       "      <td>16</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    trips_in_first_30_days signup_date  avg_rating_of_driver  avg_surge  \\\n",
       "19                       0  2014-01-02              5.000000       1.17   \n",
       "25                       1  2014-01-02              4.601559       1.00   \n",
       "37                       0  2014-01-02              4.000000       1.00   \n",
       "51                       4  2014-01-02              4.500000       1.00   \n",
       "68                       0  2014-01-01              5.000000       1.00   \n",
       "\n",
       "   last_trip_date  surge_pct  ultimate_black_user  weekday_pct  avg_dist  \\\n",
       "19     2014-06-03       33.3                    0         66.7     12.08   \n",
       "25     2014-01-03        0.0                    1        100.0     14.48   \n",
       "37     2014-04-26        0.0                    0         50.0      1.38   \n",
       "51     2014-05-02        0.0                    0         52.4      3.24   \n",
       "68     2014-06-15        0.0                    1          0.0      4.44   \n",
       "\n",
       "    avg_rating_by_driver  city_Astapor  city_King's Landing  city_Winterfell  \\\n",
       "19                   5.0             0                    1                0   \n",
       "25                   5.0             0                    0                1   \n",
       "37                   5.0             1                    0                0   \n",
       "51                   5.0             0                    0                1   \n",
       "68                   5.0             0                    1                0   \n",
       "\n",
       "    phone_Android  phone_iPhone  retained  signup_days  last_trip_days  \n",
       "19              0             1         1          180              28  \n",
       "25              0             1         0          180             179  \n",
       "37              0             1         0          180              66  \n",
       "51              1             0         0          180              60  \n",
       "68              0             1         1          181              16  "
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# create a column for number of days since signup and number of days since last trip\n",
    "df_6mo['signup_days'] = (last_trip - df_6mo['signup_date']).dt.days\n",
    "df_6mo['last_trip_days'] = (last_trip - df_6mo['last_trip_date']).dt.days\n",
    "df_6mo.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [],
   "source": [
    "# assign X and y, features and response arrays\n",
    "y = np.array(df_6mo['retained'])\n",
    "X = df_6mo.drop(columns=['signup_date', 'last_trip_date', 'retained'])\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=18)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression Score: 0.9951040391676866\n"
     ]
    }
   ],
   "source": [
    "# perform logistic regression\n",
    "log = LogisticRegression(max_iter=1000)\n",
    "log.fit(X_train, y_train)\n",
    "log_score = log.score(X_test, y_test)\n",
    "print('Logistic Regression Score: {}'.format(log_score))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[519   0]\n",
      " [  4 294]]\n"
     ]
    }
   ],
   "source": [
    "# view confusion matrix\n",
    "pred = log.predict(X_test)\n",
    "c=confusion_matrix(y_test, pred)\n",
    "print(c)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I chose logistic regression because it is quite fast, and therefore is usually my first choice of classifiers. It ended up performing extremely well, so for this exercise, I'm not going to try other classification algorithms. The accuracy score is over 99%, and the confusion matrix shows 4 false positives and 0 false negatives, indicating a very useful model.\n",
    "\n",
    "Actually, it is so accurate that I am somewhat concerned about contamination, e.g. some feature being too strongly correllated with the label y. When I look into feature coefficients in the next section, this will become more apparent."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Insights ##\n",
    "\n",
    "Feature coefficients can be used as a rough approximation of feature importance if the features are scaled first. So I will re-run the LogisticRegression with scaling and then examine feature coefficients."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [],
   "source": [
    "# scale data\n",
    "scaler = StandardScaler()\n",
    "X_train_trans = scaler.fit_transform(X_train)\n",
    "X_test_trans = scaler.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression Score: 0.9951040391676866\n"
     ]
    }
   ],
   "source": [
    "# redo logistic regression with scaled data\n",
    "log.fit(X_train_trans, y_train)\n",
    "log_score = log.score(X_test_trans, y_test)\n",
    "print('Logistic Regression Score: {}'.format(log_score))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[516   3]\n",
      " [  1 297]]\n"
     ]
    }
   ],
   "source": [
    "# view confusion matrix\n",
    "pred = log.predict(X_test_trans)\n",
    "c=confusion_matrix(y_test, pred)\n",
    "print(c)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The fit is still good, the confusion matrix is a little more even.\n",
    "\n",
    "Here are the features and their importances."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [],
   "source": [
    "importance = log.coef_[0]\n",
    "feature_dict = {}\n",
    "for i, val in enumerate(importance):\n",
    "    feature_dict[X_train.columns[i]] = val\n",
    "\n",
    "feature_df = pd.DataFrame(feature_dict, index=[0]).transpose()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_df.columns = ['feature_importance']\n",
    "feature_df = feature_df.sort_values('feature_importance')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "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_importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>last_trip_days</th>\n",
       "      <td>-10.950820</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>phone_Android</th>\n",
       "      <td>-0.161688</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>city_Astapor</th>\n",
       "      <td>-0.155200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>city_Winterfell</th>\n",
       "      <td>-0.146842</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>surge_pct</th>\n",
       "      <td>-0.124062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "      <td>-0.114476</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <td>-0.105663</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>signup_days</th>\n",
       "      <td>0.023957</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>avg_dist</th>\n",
       "      <td>0.026933</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>weekday_pct</th>\n",
       "      <td>0.117697</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>trips_in_first_30_days</th>\n",
       "      <td>0.157184</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>phone_iPhone</th>\n",
       "      <td>0.161688</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>avg_surge</th>\n",
       "      <td>0.230861</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ultimate_black_user</th>\n",
       "      <td>0.240463</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>city_King's Landing</th>\n",
       "      <td>0.368146</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                        feature_importance\n",
       "last_trip_days                  -10.950820\n",
       "phone_Android                    -0.161688\n",
       "city_Astapor                     -0.155200\n",
       "city_Winterfell                  -0.146842\n",
       "surge_pct                        -0.124062\n",
       "avg_rating_by_driver             -0.114476\n",
       "avg_rating_of_driver             -0.105663\n",
       "signup_days                       0.023957\n",
       "avg_dist                          0.026933\n",
       "weekday_pct                       0.117697\n",
       "trips_in_first_30_days            0.157184\n",
       "phone_iPhone                      0.161688\n",
       "avg_surge                         0.230861\n",
       "ultimate_black_user               0.240463\n",
       "city_King's Landing               0.368146"
      ]
     },
     "execution_count": 139,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusions ##\n",
    "\n",
    "Based on this information, some advice I might give is:\n",
    "* If you only do one thing, give a coupon or other incentive to users who have gone too long since their last ride, since this is by far the strongest predictor of retention\n",
    "* Look into whether the app is harder to use on an Android, since Android use has a negative impact on retention\n",
    "* Give coupons to users in Astapor and Winterfell\n",
    "* Incentivize taking trips in the first 30 days and potentially taking trips on weekedays, since they have a small positive impact on retention\n",
    "* Promote the Ultimate Black membership more"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}