src/B1_feature_label_generation_with_spark.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature Engineering Using Spark SQL\n",
    "Load data into Spark DataFrame, and create the features necessary for churn prediction"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load libraries, files and prepare for processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Start a Spark session, in case not invoking from CMD as 'pyspark'\n",
    "from pyspark.sql import SparkSession\n",
    "spark = SparkSession.builder.master(\"local\").appName(\"Feature Engineering\").getOrCreate()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pyspark.sql.functions as F\n",
    "import matplotlib.pyplot as plt\n",
    "plt.style.use('ggplot')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DataFrame[uid: int, event: string, song_id: double, date: timestamp]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load file\n",
    "# Took 30 secs at least on 16GB 2015 Macbook Pro\n",
    "df = spark.read.format('csv').options(header='true', inferSchema='true').load(\"../data/event_ds.csv\").cache()\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+---------+-----+-----------+-------------------+\n",
      "|      uid|event|    song_id|               date|\n",
      "+---------+-----+-----------+-------------------+\n",
      "|168540455|    P|   298250.0|2017-03-30 00:00:00|\n",
      "|168550728|    P|   285233.0|2017-03-30 00:00:00|\n",
      "|168551524|    P|  6635279.0|2017-03-30 00:00:00|\n",
      "|168543348|    P|        0.0|2017-03-30 00:00:00|\n",
      "|168550948|    P|   277286.0|2017-03-30 00:00:00|\n",
      "|168548945|    P|2.0429929E7|2017-03-30 00:00:00|\n",
      "|168550838|    P|    99018.0|2017-03-30 00:00:00|\n",
      "|168549948|    P|2.3082492E7|2017-03-30 00:00:00|\n",
      "|168547123|    P|        0.0|2017-03-30 00:00:00|\n",
      "|168551442|    P|  1474915.0|2017-03-30 00:00:00|\n",
      "|168543348|    P|        0.0|2017-03-30 00:00:00|\n",
      "|168551311|    P|  7149583.0|2017-03-30 00:00:00|\n",
      "|168551311|    P|  6359029.0|2017-03-30 00:00:00|\n",
      "|168551311|    P|  6660691.0|2017-03-30 00:00:00|\n",
      "|168550597|    P|   126530.0|2017-03-30 00:00:00|\n",
      "|168543348|    P|        0.0|2017-03-30 00:00:00|\n",
      "|168551524|    P|  6635281.0|2017-03-30 00:00:00|\n",
      "|168551406|    P|   478338.0|2017-03-30 00:00:00|\n",
      "|168551233|    P|2.3491654E7|2017-03-30 00:00:00|\n",
      "|168530975|    P|  6237732.0|2017-03-30 00:00:00|\n",
      "+---------+-----+-----------+-------------------+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Confirm that we have day-level, time-stamped data of user activity\n",
    "df.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DataFrame[uid: int, event: string, song_id: double, date: date]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Cast 'date' column into 'date' type. Especially since we don't have hour-level data anyway!\n",
    "df = df.withColumn('date', F.col('date').cast('date'))\n",
    "\n",
    "# Confirm results of typecasting\n",
    "df "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Exploration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "11523915"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Count number of events\n",
    "# In the 10 millions!\n",
    "df.count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "64527"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Count distinct users --- match with notebook 3 confirmed\n",
    "df.select('uid').distinct().count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+--------+\n",
      "|event|   count|\n",
      "+-----+--------+\n",
      "|    D|  632820|\n",
      "|    S|  791517|\n",
      "|    P|10099578|\n",
      "+-----+--------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Count each type of events\n",
    "# The vast majority are plays!\n",
    "df.groupBy('event').count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+--------+---------+\n",
      "|event|   count|  max_uid|\n",
      "+-----+--------+---------+\n",
      "|    D|  632820|169246553|\n",
      "|    S|  791517|169246553|\n",
      "|    P|10099578|169251556|\n",
      "+-----+--------+---------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# First compute the relevant, then aggregate using max/min/mean etc\n",
    "# Here we found the user who performed the most D/S/P events\n",
    "# Not useful immediately, but potentially useful for studying behavior of 'most active users'\n",
    "df.groupBy('event').agg(\n",
    "    F.count(F.col('uid')).alias('count'),\n",
    "    F.max(F.col('uid')).alias('max_uid')\n",
    ").show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### How does the amount of events changed as time progressed?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Can use df.filter, in case we're interested in a restricted time frame\n",
    "# df.filter((F.col('date')>='2017-04-01') & (F.col('date')<='2017-04-05')) \\\n",
    "#                     .groupBy('date','event').count() \\\n",
    "#                     .orderBy('date','event').show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# But since we don't have much data, there'll be no filtering\n",
    "date_count = df.groupBy('date').count().toPandas()\n",
    "\n",
    "# Sort by date for easier reading and intuitive plotting\n",
    "date_count = date_count.sort_values(by='date')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,\n",
       "        17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,\n",
       "        34, 35, 36, 37, 38, 39, 40, 41, 42, 43]),\n",
       " <a list of 44 Text xticklabel objects>)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x864 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Clear *global* decrease in the number of events, as time went on!!\n",
    "# What could have gone wrong? Is there several churning? Must explore further\n",
    "\n",
    "plt.rcParams['figure.figsize'] = (12, 12)\n",
    "plt.rcParams['font.size'] = 16\n",
    "date_count.plot.bar('date', 'count')\n",
    "\n",
    "plt.title(\"Total Number of Events, Per Day\")\n",
    "plt.ylabel(\"Frequency\")\n",
    "plt.xlabel('Date')\n",
    "plt.xticks(rotation='vertical')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### See if the above trend differs among the 3 event types\n",
    "In short, it **is** different. See eventual plot and comments below it for more details."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Count events by type and date; must cache otherwise the operations that follow take forever\n",
    "counts_by_event = df.groupBy(['event', 'date']).count().cache()\n",
    "\n",
    "# Separate Pandas DF for each event\n",
    "play_count = counts_by_event.filter(F.col('event') == 'P').toPandas()\n",
    "download_count = counts_by_event.filter(F.col('event') == 'D').toPandas()\n",
    "search_count = counts_by_event.filter(F.col('event') == 'S').toPandas()\n",
    "\n",
    "# Sort date for eventual plotting\n",
    "play_count.sort_values(by='date', inplace=True)\n",
    "download_count.sort_values(by='date', inplace=True)\n",
    "search_count.sort_values(by='date', inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Normalize by the amount on the first day\n",
    "play_count['count'] = play_count['count'] / play_count['count'].iloc[0]\n",
    "download_count['count'] = download_count['count'] / download_count['count'].iloc[0]\n",
    "search_count['count'] = search_count['count'] / search_count['count'].iloc[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([736417., 736420., 736427., 736434., 736441., 736448., 736450.,\n",
       "        736457.]), <a list of 8 Text xticklabel objects>)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x864 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Then plot!\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "play_count.plot(x='date', y='count', ax=ax, label=\"Play\", style='-+')\n",
    "download_count.plot(x='date', y='count', ax=ax, label=\"Download\", style='-+')\n",
    "search_count.plot(x='date', y='count', ax=ax, label=\"Search\", style='-+')\n",
    "\n",
    "ax.set_title(\"Normalized Count of Events, Against Time\")\n",
    "ax.set_ylabel(\"Event count, normalized on first day\")\n",
    "ax.set_xlabel(\"Date\")\n",
    "ax.legend()\n",
    "\n",
    "plt.xticks(rotation='vertical')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Business observations from graph above\n",
    "While all three events exhibit a simultaneous downwards trend *overall*, there are interesting exceptions on certain dates. For instance, see the dip in 'searches' right before 04-29; at the same time there was a small peak in 'plays'! Maybe this is not important in building a churn-prediction system, but it's worth investigating if one is to understand better the business performance of this streaming platform!\n",
    "\n",
    "Concerning the significant dips in 'play': could there be a problem with the music player, or could there be load balancing issues for streaming? Given the limited data, we cannot investigate further; but those are the questions I would ask if I were at work.\n",
    "\n",
    "All in all, once the reasons for churning are identified and the losses quantified, we can make informed decisions to mitigate further churning (e.g. by buying the right amount of hardware, not too much and not too little)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Onwards to Model Building!\n",
    "### Define Label (Churn / Not Churn)\n",
    "We need to define \"churning\" --- i.e. when a user is considered as \"inactive\". Since we only have 6 weeks of data, we will go with the simple approach of\n",
    "1. Model user acitivity on the first 4 weeks of data ('feature window')\n",
    "2. Define 'churn' as the lack of activity in the last 2 weeks ('label window')\n",
    "\n",
    "It is unacceptable to have a label window shorter than 2 weeks; after all, users may be preoccupied otherwise for a week or so, before returning to music streaming. But longer than a week? It is conceivable that they wouldn't come back. \n",
    "\n",
    "If there is sufficient data, we can actually *measure* the proportion of users who churned after two weeks of inactivity, and decide whether two weeks is a sufficiently long label window. (Definitely would try if this were to be a work project.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+----------+----------+\n",
      "| min(date)| max(date)|\n",
      "+----------+----------+\n",
      "|2017-03-30|2017-05-12|\n",
      "+----------+----------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Verify the earliest and latest dates available\n",
    "df.agg(F.min('date'), F.max('date')).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label window: 2017-04-29 ~ 2017-05-12 days: 14\n",
      "feature window: 2017-03-30 ~ 2017-04-28 days: 30\n"
     ]
    }
   ],
   "source": [
    "import datetime\n",
    "from dateutil import parser\n",
    "\n",
    "label_window_size = 14\n",
    "label_window_end_date = parser.parse('2017-05-12').date()\n",
    "label_window_start_date = label_window_end_date - datetime.timedelta(label_window_size - 1)\n",
    "print('label window:',label_window_start_date,'~',label_window_end_date,'days:',label_window_size)\n",
    "\n",
    "feature_window_size = 30\n",
    "feature_window_end_date = label_window_start_date - datetime.timedelta(1)\n",
    "feature_window_start_date = feature_window_end_date  - datetime.timedelta(feature_window_size - 1)\n",
    "print('feature window:',feature_window_start_date,'~',feature_window_end_date,'days:',feature_window_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Collect all the uid's we will model; these are users with activity within the feature window\n",
    "df_model_uid = df.filter((F.col('date')>=feature_window_start_date) & (F.col('date')<=feature_window_end_date))\\\n",
    "                    .select('uid').distinct()\n",
    "\n",
    "# Did the above users churn in the label window? (churn/active: 1/0)\n",
    "df_active_uid_in_label_window = df.filter((F.col('date')>=label_window_start_date) & (F.col('date')<=label_window_end_date))\\\n",
    "                            .select('uid').distinct().withColumn('label',F.lit(0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare label data via SQL-joins (churn label=1; active label=0)\n",
    "# Whichever user not in df_model_uid is considered churned\n",
    "df_label = df_model_uid.join(df_active_uid_in_label_window, on=['uid'], how='left')\n",
    "df_label = df_label.fillna(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+-----+\n",
      "|label|count|\n",
      "+-----+-----+\n",
      "|    1|39566|\n",
      "|    0|23511|\n",
      "+-----+-----+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Finally, print distribution of 0/1 labels\n",
    "df_label.groupBy('label').count().show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature generation\n",
    "Here we create the features that describe user activity prior to churning (or not). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Collect acitivty data in the feature window\n",
    "df_feature_window = df.filter((F.col('date')>=feature_window_start_date) & (F.col('date')<=feature_window_end_date))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "'''\n",
    "    Python list to capture all generated features\n",
    "'''\n",
    "df_feature_activity = []\n",
    "df_feature_profile = []"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Frequency features\n",
    "The number of plays, downloads, and searches in the feature window"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a function to generate frequency features for a list of time windows\n",
    "# using when().otherwise(), and list comprehension trick!\n",
    "\n",
    "def frequency_feature_generation_time_windows(df, event, time_window_list, snapshot_date):\n",
    "    \"\"\"\n",
    "        Generate frequency features for one event type and a list of time windows\n",
    "    \n",
    "        event: string 'S' 'P' or 'D'\n",
    "        time_window_list: list of integers; e.g. [1, 4, 7] refers to 1, 4, 7 days before and including snapshot_date\n",
    "        snapshot_date: Python date type\n",
    "    \"\"\"\n",
    "    \n",
    "    df_feature = df \\\n",
    "        .filter(F.col('event')==event) \\\n",
    "        .groupBy('uid') \\\n",
    "        .agg(*[F.sum(F.when((F.col('date')>=snapshot_date-datetime.timedelta(time_window-1)) & (F.col('date')<=snapshot_date),1).otherwise(0)).alias('freq_'+event+'_last_'+str(time_window)) \\\n",
    "                for time_window in time_window_list]\n",
    "            )# *[] opens list and make them comma separated\n",
    "    return df_feature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+---------+-------------+-------------+-------------+--------------+\n",
      "|      uid|freq_P_last_1|freq_P_last_3|freq_P_last_5|freq_P_last_30|\n",
      "+---------+-------------+-------------+-------------+--------------+\n",
      "|169251556|            0|            8|            8|            53|\n",
      "+---------+-------------+-------------+-------------+--------------+\n",
      "\n",
      "45\n"
     ]
    }
   ],
   "source": [
    "''' \n",
    "    As a test: generate feature for 'P' event\n",
    "    - Verified manually that for a user, that the feature is correct and valid\n",
    "    - uid choice: max in play data (copied from cell above)\n",
    "'''\n",
    "\n",
    "event = 'P'\n",
    "snapshot_date = feature_window_end_date\n",
    "tmp_df = frequency_feature_generation_time_windows(df_feature_window, event, [1,3,5,30], snapshot_date)\n",
    "tmp_df.persist()\n",
    "\n",
    "# Print the newly generated feature for uid==168551430\n",
    "# Number of plays: 179, for dates at least 5 days before feature_window_end_date\n",
    "tmp_df.filter(F.col('uid')==169251556).show()\n",
    "\n",
    "# Print the number of plays for uid==168551430 during the whole feature window; 53-8=45 indeed!\n",
    "print( df_feature_window.filter( (F.col('uid')==169251556) & (F.col('event')=='P') \\\n",
    "                         & (F.col('date') <= (snapshot_date-datetime.timedelta(4))) ).count()\n",
    "     )\n",
    "\n",
    "# Delete the temporally generated DF\n",
    "del tmp_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# generate frequency features for all event_list, time_window_list\n",
    "event_list = ['P','D','S']\n",
    "time_window_list = [1,3,7,14,30]\n",
    "snapshot_date = feature_window_end_date\n",
    "\n",
    "for event in event_list:\n",
    "    df_feature_activity.append(frequency_feature_generation_time_windows(df_feature_window,event,time_window_list,snapshot_date))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[DataFrame[uid: int, freq_P_last_1: bigint, freq_P_last_3: bigint, freq_P_last_7: bigint, freq_P_last_14: bigint, freq_P_last_30: bigint],\n",
       " DataFrame[uid: int, freq_D_last_1: bigint, freq_D_last_3: bigint, freq_D_last_7: bigint, freq_D_last_14: bigint, freq_D_last_30: bigint],\n",
       " DataFrame[uid: int, freq_S_last_1: bigint, freq_S_last_3: bigint, freq_S_last_7: bigint, freq_S_last_14: bigint, freq_S_last_30: bigint]]"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_feature_activity"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Recency features\n",
    "The number of days between the last event in the feature window, and `label_window_start_date`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "def recency_feature_generation_time_windows(df, event, snapshot_date):\n",
    "    \"\"\"\n",
    "        Generate recency features for one event type and a list of time windows\n",
    "                \n",
    "        LOGIC: convert to UNIX timestamp, calculate time difference in secs, then divide by 86400 -> days\n",
    "    \n",
    "        df: Spark DataFrame containing data from feature window\n",
    "        event: string 'S' 'P' or 'D'\n",
    "        time_window_list: list of integers; e.g. [1, 4, 7] refers to 1, 4, 7 days before and including snapshot_date\n",
    "        snapshot_date: Python date type\n",
    "    \"\"\"\n",
    "    \n",
    "    from time import mktime\n",
    "        \n",
    "    df_feature = df \\\n",
    "        .filter(F.col('event') == event) \\\n",
    "        .groupBy('uid') \\\n",
    "        .agg(( (mktime(snapshot_date.timetuple()) - F.unix_timestamp(F.max('date')))/86400 ).alias('recency_'+event))\n",
    "    \n",
    "    return df_feature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+---------+---------+\n",
      "|      uid|recency_S|\n",
      "+---------+---------+\n",
      "|169246553|      4.0|\n",
      "+---------+---------+\n",
      "\n",
      "+----------+\n",
      "| max(date)|\n",
      "+----------+\n",
      "|2017-04-25|\n",
      "+----------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "''' \n",
    "    As a test: generate feature for 'S' event\n",
    "    - Verified manually that for a user, that the feature is correct and valid\n",
    "    - uid choice: max in play data (copied from cell above)\n",
    "'''\n",
    "\n",
    "event = 'S'\n",
    "snapshot_date = label_window_start_date\n",
    "tmp_df = recency_feature_generation_time_windows(df_feature_window,event,snapshot_date)\n",
    "\n",
    "# Print the newly generated feature for uid==169251556\n",
    "tmp_df.filter(F.col('uid')==169246553).show()\n",
    "\n",
    "# Print date of the latest search in feature window, for uid==169246553\n",
    "# It is April 25 --- 4 days from label_window_start_date (April 29) indeed!\n",
    "df_feature_window.filter( (F.col('uid')==169246553) \\\n",
    "                         & (F.col('event')=='S') \\\n",
    "                         & (F.col('date') <= snapshot_date)\n",
    "                        ).agg(F.max('date')).show()\n",
    "\n",
    "# Delete the temporally generated DF\n",
    "del tmp_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# generate frequency features for all event_list, time_window_list\n",
    "snapshot_date = label_window_start_date\n",
    "event_list = ['P','D','S']\n",
    "for event in event_list:\n",
    "    df_feature_activity.append(recency_feature_generation_time_windows(df_feature_window,event,snapshot_date))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[DataFrame[uid: int, freq_P_last_1: bigint, freq_P_last_3: bigint, freq_P_last_7: bigint, freq_P_last_14: bigint, freq_P_last_30: bigint],\n",
       " DataFrame[uid: int, freq_D_last_1: bigint, freq_D_last_3: bigint, freq_D_last_7: bigint, freq_D_last_14: bigint, freq_D_last_30: bigint],\n",
       " DataFrame[uid: int, freq_S_last_1: bigint, freq_S_last_3: bigint, freq_S_last_7: bigint, freq_S_last_14: bigint, freq_S_last_30: bigint],\n",
       " DataFrame[uid: int, recency_P: double],\n",
       " DataFrame[uid: int, recency_D: double],\n",
       " DataFrame[uid: int, recency_S: double]]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_feature_activity"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Profile features\n",
    "Given the available data, 'profile' refers to the device (iOS/android) that generated logged events"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+---------+------+---------+----------+---------+-----------+\n",
      "|      uid|device|  song_id|      date|play_time|song_length|\n",
      "+---------+------+---------+----------+---------+-----------+\n",
      "|168540455|    ar| 298250.0|2017-03-30|    189.0|      190.0|\n",
      "|168550728|    ar| 285233.0|2017-03-30|    282.0|      282.0|\n",
      "|168551524|    ar|6635279.0|2017-03-30|      1.0|      231.0|\n",
      "|168543348|    ar|      0.0|2017-03-30|     27.0|       63.0|\n",
      "|168550948|    ar| 277286.0|2017-03-30|    254.0|      254.0|\n",
      "+---------+------+---------+----------+---------+-----------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Load play records\n",
    "df_play = spark.read.csv('../data/play_ds_cleaned.csv', header=True)\n",
    "df_play.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Filter down to obtain data in feature window\n",
    "# Then, extract unique PAIRS of uid,device\n",
    "df_play_feature_window = df_play.filter((F.col('date')>=feature_window_start_date) & (F.col('date')<=feature_window_end_date))\n",
    "df_device_tmp = df_play_feature_window.select('uid','device').distinct()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+-----+\n",
      "|device|count|\n",
      "+------+-----+\n",
      "|    mc|    1|\n",
      "|    ar|54925|\n",
      "|    ip| 7965|\n",
      "+------+-----+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Print the number of instances per device type\n",
    "df_device_tmp.groupBy('device').count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unique number of users:  65417\n",
      "Unique number of user/device pairs:  62891\n"
     ]
    }
   ],
   "source": [
    "# There ARE users who accessed the service with >1 devices\n",
    "# df.select('uid').distinct().count() is 65417\n",
    "print(\"Unique number of users: \", 65417)\n",
    "print(\"Unique number of user/device pairs: \", df_device_tmp.distinct().count())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "13"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# There aren't actually THAT many users who have played songs on multiple devices though!\n",
    "# Explanaton of discrepancy: df_profile_tmp is built from play data only\n",
    "# ...whereas df_label is built from the entire event set\n",
    "\n",
    "multi_device_user = df_device_tmp.groupby('uid').count().filter(F.col('count') > 1)\n",
    "multi_device_user.count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----------+-----+\n",
      "|device_type|count|\n",
      "+-----------+-----+\n",
      "|          1| 7965|\n",
      "|          2|54926|\n",
      "+-----------+-----+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Encode device types into numbers. iOS == 1, android == 2\n",
    "# There is a single instance of device 'mc', which we arbitarily grouped with 2\n",
    "\n",
    "df_device_tmp = df_device_tmp.withColumn('device_type', \\\n",
    "                                           F.when(F.col('device')=='ip',1).otherwise(2))\n",
    "\n",
    "# Should match with count from a few cells above\n",
    "df_device_tmp.groupBy('device_type').count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Build Spark DataFrame that matches label (from df_label) with device\n",
    "# Only keep those uid's that have a label; hence left join\n",
    "\n",
    "df_device = df_label.select('uid').join(df_device_tmp.select('uid','device_type'),on='uid',how='left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# See above: null values in df_profile because of mismatches in uid\n",
    "# Fill as '3', stand-in for 'unknown'\n",
    "df_device = df_device.fillna({'device_type':3})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+---------+-----+\n",
      "|      uid|count|\n",
      "+---------+-----+\n",
      "|168653359|    2|\n",
      "|168165762|    2|\n",
      "|169089329|    2|\n",
      "|168576848|    2|\n",
      "|  3647844|    2|\n",
      "|168576854|    2|\n",
      "|168577159|    2|\n",
      "|168653255|    2|\n",
      "|168651933|    2|\n",
      "|168652896|    2|\n",
      "|168653231|    2|\n",
      "|168577178|    2|\n",
      "|168652781|    2|\n",
      "+---------+-----+\n",
      "\n",
      "+---------+-----------+\n",
      "|      uid|device_type|\n",
      "+---------+-----------+\n",
      "|168165762|          2|\n",
      "|168165762|          1|\n",
      "+---------+-----------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Confirm that df_profile still records the multiple devices owned by certain users\n",
    "df_device.groupBy('uid').count().filter(F.col('count') > 1).show()\n",
    "\n",
    "# Print for a particular user just displayed\n",
    "df_device.filter(F.col('uid')==168165762).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Finally, append device info to feature Python list\n",
    "df_feature_profile.append(df_device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Total play time\n",
    "The time (secs) that users have spent playing songs, in various time windows"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Cast to date type, for convenient windowing\n",
    "df_play = df_play.withColumn('date', F.col('date').cast('date'))\n",
    "\n",
    "# Cast to float type, as intended (e.g. see notebook 3.5)\n",
    "df_play = df_play.withColumn('play_time', F.col('play_time').cast('float'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of df_play entries with negative play_time:  0\n"
     ]
    }
   ],
   "source": [
    "# Can you generate total song play time features (using play_ds data) for different time window\n",
    "# using play data (need to clean play time first, play time may be negative in data)\n",
    "print(\"Number of df_play entries with negative play_time: \", \\\n",
    "      df_play.select('play_time').filter(F.col('play_time') < 0).count())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fix negative play_time by setting to 0 --- but none found\n",
    "# i.e. Assume it's a system glitch --- songs weren't actually played\n",
    "# df_play_tmp = df_play.withColumn('play_time', \\\n",
    "#                                  F.when(df_play['play_time']<0, 0).otherwise(df_play['play_time']))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_play_tmp = df_play"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "def play_time_feature_generation_time_windows(df, time_window_list, snapshot_date):\n",
    "    \"\"\"\n",
    "        Compute the total play time (secs) within a time window, before and including the snapshot date\n",
    "        \n",
    "        df: Spark DataFrame containing play data *only*\n",
    "        time_window_list: list of integers; e.g. [1, 4, 7] refers to 1, 4, 7 days before and including snapshot_date\n",
    "        snapshot_date: Python date type\n",
    "    \"\"\"\n",
    "    \n",
    "    df_feature = df \\\n",
    "        .groupBy('uid') \\\n",
    "        .agg(*[F.sum( \\\n",
    "                     F.when( (F.col('date') >= snapshot_date-datetime.timedelta(time_window-1)) & (F.col('date')<=snapshot_date), F.col('play_time')).otherwise(0))\\\n",
    "               .alias('play_time_last_'+str(time_window)) for time_window in time_window_list])\n",
    "    \n",
    "    return df_feature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DataFrame[uid: string, device: string, song_id: string, date: date, play_time: float, song_length: string]"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_play_tmp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------+\n",
      "|max(play_time)|\n",
      "+--------------+\n",
      "|       65500.0|\n",
      "+--------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_play_tmp.agg(F.max('play_time')).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generate play time features and append df_feature_activity\n",
    "time_window_list = [1,3,7,14,30]\n",
    "snapshot_date = feature_window_end_date\n",
    "df_feature_activity.append(play_time_feature_generation_time_windows(df_play_tmp, time_window_list, snapshot_date))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fancier frequency features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Can you generate counts of songs play 80% of their song length (using play_ds data) for different time window\n",
    "# using play data (need to clean play time and song length first, \n",
    "# play time may be negative in data, song length may be zeros)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Form Training Matrix and Save to File"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "def join_feature_data(df_master,df_feature_list):\n",
    "    for df_feature in df_feature_list:\n",
    "        df_master = df_master.join(df_feature,on='uid',how='left')\n",
    "        df_master.persist() # uncomment if there are too many DFs to joins\n",
    "    return df_master"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "# join all behavior features\n",
    "df_model_final = join_feature_data(df_label, df_feature_activity)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['uid',\n",
       " 'label',\n",
       " 'freq_P_last_1',\n",
       " 'freq_P_last_3',\n",
       " 'freq_P_last_7',\n",
       " 'freq_P_last_14',\n",
       " 'freq_P_last_30',\n",
       " 'freq_D_last_1',\n",
       " 'freq_D_last_3',\n",
       " 'freq_D_last_7',\n",
       " 'freq_D_last_14',\n",
       " 'freq_D_last_30',\n",
       " 'freq_S_last_1',\n",
       " 'freq_S_last_3',\n",
       " 'freq_S_last_7',\n",
       " 'freq_S_last_14',\n",
       " 'freq_S_last_30',\n",
       " 'recency_P',\n",
       " 'recency_D',\n",
       " 'recency_S',\n",
       " 'play_time_last_1',\n",
       " 'play_time_last_3',\n",
       " 'play_time_last_7',\n",
       " 'play_time_last_14',\n",
       " 'play_time_last_30']"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_model_final.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "# join all profile features\n",
    "df_model_final = join_feature_data(df_model_final, df_feature_profile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_model_final.fillna(0).toPandas().to_csv('../data/df_model_final.csv',index=False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}