part3/train-model/results-of-logistic-regression.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<frozen importlib._bootstrap>:228: RuntimeWarning: scipy._lib.messagestream.MessageStream size changed, may indicate binary incompatibility. Expected 56 from C header, got 64 from PyObject\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from matplotlib import pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook offers an interpretation of the results of the logistic regression analysis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "col_order = ['group_metrics', 'group_metric_offset', 'weight', 'retain_impact']\n",
    "logreg_summary_df = pd.read_csv('../../output/socialnet7_dataset2_logreg_summary.csv')\n",
    "logreg_summary_df = logreg_summary_df[col_order]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`weight` - a positive weight indicates that a metric (or group) is associated with increased engagement.\n",
    "\n",
    "`retain_impact` - represents how much difference in the retention probability it would make to be one standard deviation above average for this metric (or group), assuming a customer was average in all other aspects.\n",
    "\n",
    "If the retention impact for a metric is 2%, a customer who is one standard deviation above average on that metric and average in all the other metrics has a forecast retention probability 2% higher than the average retention probability.\n",
    "\n",
    "The results for the model offset are included. The `weight` for the offset is not a weight, but just the amount of the offset. The `retain_impact` of the offset is the retention probability forecast for a perfectly average customer (i.e. one with zero scores in all metrics).\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>group_metrics</th>\n",
       "      <th>group_metric_offset</th>\n",
       "      <th>weight</th>\n",
       "      <th>retain_impact</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>(baseline)</td>\n",
       "      <td>offset</td>\n",
       "      <td>3.534486</td>\n",
       "      <td>0.971653</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>message_per_month|reply_per_month</td>\n",
       "      <td>metric_group_3</td>\n",
       "      <td>0.677812</td>\n",
       "      <td>0.013751</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>adview_per_month|like_per_month|newfriend_per_month|post_per_message|post_per_month</td>\n",
       "      <td>metric_group_1</td>\n",
       "      <td>0.610598</td>\n",
       "      <td>0.012752</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>newfriend_pcnt_chng</td>\n",
       "      <td>newfriend_pcnt_chng</td>\n",
       "      <td>0.175516</td>\n",
       "      <td>0.004454</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>dislike_pcnt</td>\n",
       "      <td>dislike_pcnt</td>\n",
       "      <td>-0.002006</td>\n",
       "      <td>-0.000055</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>reply_per_message</td>\n",
       "      <td>reply_per_message</td>\n",
       "      <td>-0.026286</td>\n",
       "      <td>-0.000733</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>unfriend_per_month|unfriend_per_newfriend</td>\n",
       "      <td>metric_group_2</td>\n",
       "      <td>-0.104359</td>\n",
       "      <td>-0.003020</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>dislike_per_month</td>\n",
       "      <td>dislike_per_month</td>\n",
       "      <td>-0.112772</td>\n",
       "      <td>-0.003277</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>days_since_newfriend</td>\n",
       "      <td>days_since_newfriend</td>\n",
       "      <td>-0.147217</td>\n",
       "      <td>-0.004349</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>adview_per_post</td>\n",
       "      <td>adview_per_post</td>\n",
       "      <td>-0.318860</td>\n",
       "      <td>-0.010235</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                         group_metrics  \\\n",
       "9                                                                           (baseline)   \n",
       "2                                                    message_per_month|reply_per_month   \n",
       "0  adview_per_month|like_per_month|newfriend_per_month|post_per_message|post_per_month   \n",
       "7                                                                  newfriend_pcnt_chng   \n",
       "6                                                                         dislike_pcnt   \n",
       "5                                                                    reply_per_message   \n",
       "1                                            unfriend_per_month|unfriend_per_newfriend   \n",
       "3                                                                    dislike_per_month   \n",
       "8                                                                 days_since_newfriend   \n",
       "4                                                                      adview_per_post   \n",
       "\n",
       "    group_metric_offset    weight  retain_impact  \n",
       "9                offset  3.534486       0.971653  \n",
       "2        metric_group_3  0.677812       0.013751  \n",
       "0        metric_group_1  0.610598       0.012752  \n",
       "7   newfriend_pcnt_chng  0.175516       0.004454  \n",
       "6          dislike_pcnt -0.002006      -0.000055  \n",
       "5     reply_per_message -0.026286      -0.000733  \n",
       "1        metric_group_2 -0.104359      -0.003020  \n",
       "3     dislike_per_month -0.112772      -0.003277  \n",
       "8  days_since_newfriend -0.147217      -0.004349  \n",
       "4       adview_per_post -0.318860      -0.010235  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.set_option('display.max_colwidth', None)\n",
    "logreg_summary_df.sort_values(by=['retain_impact'], ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plotting the impacts on retention probability as a bar chart makes them easier to interpret. We also create a set of labels for groups to explain the theme of the metric:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "label_dict = {\n",
    "    'metric_group_1': 'Posting and making friends',\n",
    "    'metric_group_2': 'Unfriending',\n",
    "    'metric_group_3': 'Messaging',\n",
    "    'newfriend_pcnt_chng': 'Increase in new friends',\n",
    "    'dislike_pcnt': 'More dislikes than likes',\n",
    "    'reply_per_message': 'Replies per message',\n",
    "    'adview_per_post': 'Seeing lots of ads',\n",
    "    'days_since_newfriend': 'Time since new friend',\n",
    "    'dislike_per_month': 'Dislikes per month'\n",
    "}\n",
    "\n",
    "rows_to_drop = logreg_summary_df[logreg_summary_df['group_metric_offset'] == 'offset']\n",
    "sorted_df = (\n",
    "    logreg_summary_df.drop(rows_to_drop.index)\n",
    "    .sort_values(by='retain_impact', ascending=False)\n",
    "    .assign(retain_impact=lambda x: x['retain_impact'] * 100) # convert to percentage\n",
    ")\n",
    "# Add a column of labels describing the theme of the metric\n",
    "sorted_df['metric_theme'] = sorted_df['group_metric_offset'].map(label_dict)\n",
    "\n",
    "sns.barplot(data=sorted_df, x='metric_theme', y='retain_impact')\n",
    "plt.title('Impact on retention probability of being above average (by one standard deviation)')\n",
    "plt.xlabel('')\n",
    "plt.ylabel('Impact %')\n",
    "plt.xticks(rotation=90, fontsize=8)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "These results would suggest that _being above average_ in messaging activity, posting and continuing to make friends have the biggest positive impact on retention, while _being above average_ seeing lots of ads, going long periods between making new friends and disliking content have the biggest negative impact on retention.\n",
    "\n",
    "Interestingly, the ratio of likes to dislikes has a negligible impact on churn/retention.\n",
    "\n",
    "It's also important to note that:\n",
    "- Being below average will have an approximately equal and opposite effect.\n",
    "- If a customer is multiple standard deviations above average, there are diminishing returns, meaning that each additional standard deviation above average has less impact on the churn or retention probability.\n",
    "- The same diminishing returns goes for being above average in multiple respects:\n",
    "the combined churn probability reduction will be lower than the sum of the quoted retention probability impacts."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before going further, we should check that the model is calibrated. A good check of model calibration is to compare the historical churn probabilities predicted by the model to the summary statistics of the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "churn_prob     0.047886\n",
       "retain_prob    0.952114\n",
       "dtype: float64"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions_df = pd.read_csv('../../output/socialnet7_dataset2_predictions.csv')\n",
    "\n",
    "# Show the mean churn and retention probabilities predicted by the model\n",
    "predictions_df[['churn_prob', 'retain_prob']].mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>count</th>\n",
       "      <th>nonzero</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>skew</th>\n",
       "      <th>min</th>\n",
       "      <th>1pct</th>\n",
       "      <th>25pct</th>\n",
       "      <th>50pct</th>\n",
       "      <th>75pct</th>\n",
       "      <th>99pct</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>is_churn</td>\n",
       "      <td>25143</td>\n",
       "      <td>0.047846</td>\n",
       "      <td>0.047846</td>\n",
       "      <td>0.213445</td>\n",
       "      <td>4.237056</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Unnamed: 0  count   nonzero      mean       std      skew  min  1pct  25pct  \\\n",
       "0   is_churn  25143  0.047846  0.047846  0.213445  4.237056    0     0    0.0   \n",
       "\n",
       "   50pct  75pct  99pct  max  \n",
       "0    0.0    0.0    1.0  1.0  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "summary_stats_df = pd.read_csv('../../output/socialnet7_dataset2_summarystats.csv')\n",
    "\n",
    "# Grab the summary stat for just churn\n",
    "summary_stats_df.loc[summary_stats_df.iloc[:, 0] == 'is_churn']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that the mean churn probability predicted by the model is very close to the churn rate of the sample dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "churn",
   "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": "3.9.16"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}