Sparkify.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sparkify Project Workspace\n",
    "This workspace contains a tiny subset (128MB) of the full dataset available (12GB). Feel free to use this workspace to build your project, or to explore a smaller subset with Spark before deploying your cluster on the cloud. Instructions for setting up your Spark cluster is included in the last lesson of the Extracurricular Spark Course content.\n",
    "\n",
    "You can follow the steps below to guide your data analysis and model building portion of this project."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import libraries\n",
    "from pyspark.sql import SparkSession\n",
    "from pyspark.sql import functions as F\n",
    "from pyspark.sql import Window\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from pyspark.ml.feature import VectorAssembler, StandardScaler, MinMaxScaler\n",
    "from pyspark.ml import Pipeline\n",
    "from pyspark.ml.evaluation import MulticlassClassificationEvaluator\n",
    "from pyspark.ml.tuning import CrossValidator, ParamGridBuilder\n",
    "from pyspark.ml.classification import LogisticRegression, RandomForestClassifier, GBTClassifier\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create a Spark session\n",
    "spark = SparkSession.builder \\\n",
    "    .master(\"local\") \\\n",
    "    .appName(\"Sparkify\") \\\n",
    "    .getOrCreate()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>pre { white-space: pre !important; }</style>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# For better dataframe display purpose\n",
    "from IPython.core.display import HTML\n",
    "display(HTML(\"<style>pre { white-space: pre !important; }</style>\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load and Clean Dataset\n",
    "In this workspace, the mini-dataset file is `mini_sparkify_event_data.json`. Load and clean the dataset, checking for invalid or missing data - for example, records without userids or sessionids. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Row(artist='Martha Tilston', auth='Logged In', firstName='Colin', gender='M', itemInSession=50, lastName='Freeman', length=277.89016, level='paid', location='Bakersfield, CA', method='PUT', page='NextSong', registration=1538173362000, sessionId=29, song='Rockpools', status=200, ts=1538352117000, userAgent='Mozilla/5.0 (Windows NT 6.1; WOW64; rv:31.0) Gecko/20100101 Firefox/31.0', userId='30')"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "path = 'mini_sparkify_event_data.json'\n",
    "df = spark.read.json(path)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+----------------+---------+---------+------+-------------+--------+---------+-----+--------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+\n",
      "|          artist|     auth|firstName|gender|itemInSession|lastName|   length|level|            location|method|    page| registration|sessionId|                song|status|           ts|           userAgent|userId|\n",
      "+----------------+---------+---------+------+-------------+--------+---------+-----+--------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+\n",
      "|  Martha Tilston|Logged In|    Colin|     M|           50| Freeman|277.89016| paid|     Bakersfield, CA|   PUT|NextSong|1538173362000|       29|           Rockpools|   200|1538352117000|Mozilla/5.0 (Wind...|    30|\n",
      "|Five Iron Frenzy|Logged In|    Micah|     M|           79|    Long|236.09424| free|Boston-Cambridge-...|   PUT|NextSong|1538331630000|        8|              Canada|   200|1538352180000|\"Mozilla/5.0 (Win...|     9|\n",
      "|    Adam Lambert|Logged In|    Colin|     M|           51| Freeman| 282.8273| paid|     Bakersfield, CA|   PUT|NextSong|1538173362000|       29|   Time For Miracles|   200|1538352394000|Mozilla/5.0 (Wind...|    30|\n",
      "|          Enigma|Logged In|    Micah|     M|           80|    Long|262.71302| free|Boston-Cambridge-...|   PUT|NextSong|1538331630000|        8|Knocking On Forbi...|   200|1538352416000|\"Mozilla/5.0 (Win...|     9|\n",
      "|       Daft Punk|Logged In|    Colin|     M|           52| Freeman|223.60771| paid|     Bakersfield, CA|   PUT|NextSong|1538173362000|       29|Harder Better Fas...|   200|1538352676000|Mozilla/5.0 (Wind...|    30|\n",
      "+----------------+---------+---------+------+-------------+--------+---------+-----+--------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- artist: string (nullable = true)\n",
      " |-- auth: string (nullable = true)\n",
      " |-- firstName: string (nullable = true)\n",
      " |-- gender: string (nullable = true)\n",
      " |-- itemInSession: long (nullable = true)\n",
      " |-- lastName: string (nullable = true)\n",
      " |-- length: double (nullable = true)\n",
      " |-- level: string (nullable = true)\n",
      " |-- location: string (nullable = true)\n",
      " |-- method: string (nullable = true)\n",
      " |-- page: string (nullable = true)\n",
      " |-- registration: long (nullable = true)\n",
      " |-- sessionId: long (nullable = true)\n",
      " |-- song: string (nullable = true)\n",
      " |-- status: long (nullable = true)\n",
      " |-- ts: long (nullable = true)\n",
      " |-- userAgent: string (nullable = true)\n",
      " |-- userId: string (nullable = true)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.printSchema()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "286500"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Clean Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-------+------------------+----------+---------+------+------------------+--------+-----------------+------+-----------------+------+-------+--------------------+-----------------+--------------------+------------------+--------------------+--------------------+-----------------+\n",
      "|summary|            artist|      auth|firstName|gender|     itemInSession|lastName|           length| level|         location|method|   page|        registration|        sessionId|                song|            status|                  ts|           userAgent|           userId|\n",
      "+-------+------------------+----------+---------+------+------------------+--------+-----------------+------+-----------------+------+-------+--------------------+-----------------+--------------------+------------------+--------------------+--------------------+-----------------+\n",
      "|  count|            228108|    286500|   278154|278154|            286500|  278154|           228108|286500|           278154|286500| 286500|              278154|           286500|              228108|            286500|              286500|              278154|           286500|\n",
      "|   mean| 551.0852017937219|      null|     null|  null|114.41421291448516|    null|249.1171819778458|  null|             null|  null|   null|1.535358834084427...|1041.526554973822|            Infinity|210.05459685863875|1.540956889810483...|                null|59682.02278593872|\n",
      "| stddev|1217.7693079161374|      null|     null|  null|129.76726201140994|    null|99.23517921058361|  null|             null|  null|   null| 3.291321616327586E9|726.7762634630741|                 NaN| 31.50507848842214|1.5075439608226302E9|                null|109091.9499991047|\n",
      "|    min|               !!!| Cancelled| Adelaida|     F|                 0|   Adams|          0.78322|  free|       Albany, OR|   GET|  About|       1521380675000|                1|\u001c",
      "ÃÂg ÃÂtti Gr...|               200|       1538352117000|\"Mozilla/5.0 (Mac...|                 |\n",
      "|    max| ÃÂlafur Arnalds|Logged Out|   Zyonna|     M|              1321|  Wright|       3024.66567|  paid|Winston-Salem, NC|   PUT|Upgrade|       1543247354000|             2474|ÃÂau hafa slopp...|               404|       1543799476000|Mozilla/5.0 (comp...|               99|\n",
      "+-------+------------------+----------+---------+------+------------------+--------+-----------------+------+-----------------+------+-------+--------------------+-----------------+--------------------+------------------+--------------------+--------------------+-----------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.describe().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(auth='Logged Out'),\n",
       " Row(auth='Cancelled'),\n",
       " Row(auth='Guest'),\n",
       " Row(auth='Logged In')]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.select('auth').distinct().collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(level='free'), Row(level='paid')]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.select('level').distinct().collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(page='Cancel'),\n",
       " Row(page='Submit Downgrade'),\n",
       " Row(page='Thumbs Down'),\n",
       " Row(page='Home'),\n",
       " Row(page='Downgrade'),\n",
       " Row(page='Roll Advert'),\n",
       " Row(page='Logout'),\n",
       " Row(page='Save Settings'),\n",
       " Row(page='Cancellation Confirmation'),\n",
       " Row(page='About'),\n",
       " Row(page='Submit Registration'),\n",
       " Row(page='Settings'),\n",
       " Row(page='Login'),\n",
       " Row(page='Register'),\n",
       " Row(page='Add to Playlist'),\n",
       " Row(page='Add Friend'),\n",
       " Row(page='NextSong'),\n",
       " Row(page='Thumbs Up'),\n",
       " Row(page='Help'),\n",
       " Row(page='Upgrade'),\n",
       " Row(page='Error'),\n",
       " Row(page='Submit Upgrade')]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.select('page').distinct().collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(method='PUT'), Row(method='GET')]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.select('method').distinct().collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(level='free'), Row(level='paid')]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.select('level').distinct().collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "226"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.select('userId').distinct().count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2354"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.select('sessionId').distinct().count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+-----+------+---+---------+------+\n",
      "|artist|auth|firstName|gender|itemInSession|lastName|length|level|location|method|page|registration|sessionId| song|status| ts|userAgent|userId|\n",
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+-----+------+---+---------+------+\n",
      "| 58392|   0|     8346|  8346|            0|    8346| 58392|    0|    8346|     0|   0|        8346|        0|58392|     0|  0|     8346|     0|\n",
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+-----+------+---+---------+------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.select([F.count(F.when(F.col(c).isNull(), c)).alias(c) for c in df.columns]).show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that number of null value in `artist, length, song` is the same."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+----+------+---+---------+------+\n",
      "|artist|auth|firstName|gender|itemInSession|lastName|length|level|location|method|page|registration|sessionId|song|status| ts|userAgent|userId|\n",
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+----+------+---+---------+------+\n",
      "|     0|   0|        0|     0|            0|       0|     0|    0|       0|     0|   0|           0|        0|   0|     0|  0|        0|  8346|\n",
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+----+------+---+---------+------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.select([F.count(F.when(F.col(c) == \"\", c)).alias(c) for c in df.columns]).show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There are many records of `userId` which is a empty string."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_userId = df.filter(F.col(\"userId\") == \"\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+----------+---------+------+-------------+--------+------+-----+--------+------+-----+------------+---------+----+------+-------------+---------+------+\n",
      "|artist|      auth|firstName|gender|itemInSession|lastName|length|level|location|method| page|registration|sessionId|song|status|           ts|userAgent|userId|\n",
      "+------+----------+---------+------+-------------+--------+------+-----+--------+------+-----+------------+---------+----+------+-------------+---------+------+\n",
      "|  null|Logged Out|     null|  null|          100|    null|  null| free|    null|   GET| Home|        null|        8|null|   200|1538355745000|     null|      |\n",
      "|  null|Logged Out|     null|  null|          101|    null|  null| free|    null|   GET| Help|        null|        8|null|   200|1538355807000|     null|      |\n",
      "|  null|Logged Out|     null|  null|          102|    null|  null| free|    null|   GET| Home|        null|        8|null|   200|1538355841000|     null|      |\n",
      "|  null|Logged Out|     null|  null|          103|    null|  null| free|    null|   PUT|Login|        null|        8|null|   307|1538355842000|     null|      |\n",
      "|  null|Logged Out|     null|  null|            2|    null|  null| free|    null|   GET| Home|        null|      240|null|   200|1538356678000|     null|      |\n",
      "+------+----------+---------+------+-------------+--------+------+-----+--------+------+-----+------------+---------+----+------+-------------+---------+------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_userId.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(page='Home'),\n",
       " Row(page='About'),\n",
       " Row(page='Submit Registration'),\n",
       " Row(page='Login'),\n",
       " Row(page='Register'),\n",
       " Row(page='Help'),\n",
       " Row(page='Error')]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_userId.select('page').distinct().collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(auth='Logged Out'), Row(auth='Guest')]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_userId.select('auth').distinct().collect()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As empty `userId` means the user is not log in to the application, we don't need to keep these data as it does not relevant to customer churn."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.filter(F.col(\"userId\") != \"\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Check `artist` data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_artist = df.filter(F.col(\"artist\").isNull())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+---------------+-------------+---------+----+------+-------------+--------------------+------+\n",
      "|artist|     auth|firstName|gender|itemInSession|lastName|length|level|            location|method|           page| registration|sessionId|song|status|           ts|           userAgent|userId|\n",
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+---------------+-------------+---------+----+------+-------------+--------------------+------+\n",
      "|  null|Logged In|    Colin|     M|           54| Freeman|  null| paid|     Bakersfield, CA|   PUT|Add to Playlist|1538173362000|       29|null|   200|1538352905000|Mozilla/5.0 (Wind...|    30|\n",
      "|  null|Logged In|    Micah|     M|           84|    Long|  null| free|Boston-Cambridge-...|   GET|    Roll Advert|1538331630000|        8|null|   200|1538353150000|\"Mozilla/5.0 (Win...|     9|\n",
      "|  null|Logged In|    Micah|     M|           86|    Long|  null| free|Boston-Cambridge-...|   PUT|      Thumbs Up|1538331630000|        8|null|   307|1538353376000|\"Mozilla/5.0 (Win...|     9|\n",
      "|  null|Logged In|    Alexi|     F|            4|  Warren|  null| paid|Spokane-Spokane V...|   GET|      Downgrade|1532482662000|       53|null|   200|1538354749000|Mozilla/5.0 (Wind...|    54|\n",
      "|  null|Logged In|    Alexi|     F|            7|  Warren|  null| paid|Spokane-Spokane V...|   PUT|      Thumbs Up|1532482662000|       53|null|   307|1538355255000|Mozilla/5.0 (Wind...|    54|\n",
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+---------------+-------------+---------+----+------+-------------+--------------------+------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_artist.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(page='Cancel'),\n",
       " Row(page='Submit Downgrade'),\n",
       " Row(page='Thumbs Down'),\n",
       " Row(page='Home'),\n",
       " Row(page='Downgrade'),\n",
       " Row(page='Roll Advert'),\n",
       " Row(page='Logout'),\n",
       " Row(page='Save Settings'),\n",
       " Row(page='Cancellation Confirmation'),\n",
       " Row(page='About'),\n",
       " Row(page='Settings'),\n",
       " Row(page='Add to Playlist'),\n",
       " Row(page='Add Friend'),\n",
       " Row(page='Thumbs Up'),\n",
       " Row(page='Help'),\n",
       " Row(page='Upgrade'),\n",
       " Row(page='Error'),\n",
       " Row(page='Submit Upgrade')]"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_artist.select('page').distinct().collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+-----+------+---+---------+------+\n",
      "|artist|auth|firstName|gender|itemInSession|lastName|length|level|location|method|page|registration|sessionId| song|status| ts|userAgent|userId|\n",
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+-----+------+---+---------+------+\n",
      "| 50046|   0|        0|     0|            0|       0| 50046|    0|       0|     0|   0|           0|        0|50046|     0|  0|        0|     0|\n",
      "+------+----+---------+------+-------------+--------+------+-----+--------+------+----+------------+---------+-----+------+---+---------+------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_artist.select([F.count(F.when(F.col(c).isNull(), c)).alias(c) for c in df_artist.columns]).show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We cannot remove record that its `artist` is null because it is a normal behavior when the user interact with the application, and this include \"Upgrade\" or \"Cancellation\" activity, etc.\n",
    "\n",
    "Number of null value in `df_artist` is the same with `df`."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exploratory Data Analysis\n",
    "When you're working with the full dataset, perform EDA by loading a small subset of the data and doing basic manipulations within Spark. In this workspace, you are already provided a small subset of data you can explore.\n",
    "\n",
    "### Define Churn\n",
    "\n",
    "Once you've done some preliminary analysis, create a column `Churn` to use as the label for your model. I suggest using the `Cancellation Confirmation` events to define your churn, which happen for both paid and free users. As a bonus task, you can also look into the `Downgrade` events.\n",
    "\n",
    "### Explore Data\n",
    "Once you've defined churn, perform some exploratory data analysis to observe the behavior for users who stayed vs users who churned. You can start by exploring aggregates on these two groups of users, observing how much of a specific action they experienced per a certain time unit or number of songs played."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.withColumn(\n",
    "    \"Churn\",\n",
    "    F.when(F.col(\"page\") == \"Cancellation Confirmation\", 1).otherwise(0)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+----------------+---------+---------+------+-------------+--------+---------+-----+--------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+\n",
      "|          artist|     auth|firstName|gender|itemInSession|lastName|   length|level|            location|method|    page| registration|sessionId|                song|status|           ts|           userAgent|userId|Churn|\n",
      "+----------------+---------+---------+------+-------------+--------+---------+-----+--------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+\n",
      "|  Martha Tilston|Logged In|    Colin|     M|           50| Freeman|277.89016| paid|     Bakersfield, CA|   PUT|NextSong|1538173362000|       29|           Rockpools|   200|1538352117000|Mozilla/5.0 (Wind...|    30|    0|\n",
      "|Five Iron Frenzy|Logged In|    Micah|     M|           79|    Long|236.09424| free|Boston-Cambridge-...|   PUT|NextSong|1538331630000|        8|              Canada|   200|1538352180000|\"Mozilla/5.0 (Win...|     9|    0|\n",
      "|    Adam Lambert|Logged In|    Colin|     M|           51| Freeman| 282.8273| paid|     Bakersfield, CA|   PUT|NextSong|1538173362000|       29|   Time For Miracles|   200|1538352394000|Mozilla/5.0 (Wind...|    30|    0|\n",
      "|          Enigma|Logged In|    Micah|     M|           80|    Long|262.71302| free|Boston-Cambridge-...|   PUT|NextSong|1538331630000|        8|Knocking On Forbi...|   200|1538352416000|\"Mozilla/5.0 (Win...|     9|    0|\n",
      "|       Daft Punk|Logged In|    Colin|     M|           52| Freeman|223.60771| paid|     Bakersfield, CA|   PUT|NextSong|1538173362000|       29|Harder Better Fas...|   200|1538352676000|Mozilla/5.0 (Wind...|    30|    0|\n",
      "+----------------+---------+---------+------+-------------+--------+---------+-----+--------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+------+\n",
      "|Churn| count|\n",
      "+-----+------+\n",
      "|    1|    52|\n",
      "|    0|278102|\n",
      "+-----+------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.groupBy(\"Churn\").count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "225"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.select('userId').distinct().count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Churn rate: 23.11111111111111\n"
     ]
    }
   ],
   "source": [
    "print(\"Churn rate:\", 52/225*100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "window = Window.partitionBy(\"userId\").orderBy(\"ts\").rowsBetween(Window.unboundedPreceding, Window.currentRow)\n",
    "df = df.withColumn(\"num_of_churn\", F.sum(\"Churn\").over(window))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+--------------------+-------------+---------+----+------+-------------+--------------------+------+-----+------------+\n",
      "|artist|     auth|firstName|gender|itemInSession|lastName|length|level|            location|method|                page| registration|sessionId|song|status|           ts|           userAgent|userId|Churn|num_of_churn|\n",
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+--------------------+-------------+---------+----+------+-------------+--------------------+------+-----+------------+\n",
      "|  null|Cancelled|    Mason|     M|           10|    Hart|  null| free|  Corpus Christi, TX|   GET|Cancellation Conf...|1533157139000|      174|null|   200|1539318918000|\"Mozilla/5.0 (Mac...|   125|    1|           1|\n",
      "|  null|Cancelled|    Ethan|     M|          176| Johnson|  null| paid|Lexington-Fayette...|   GET|Cancellation Conf...|1538080987000|      934|null|   200|1539761972000|\"Mozilla/5.0 (Win...|    51|    1|           1|\n",
      "|  null|Cancelled|    Alexi|     F|           42|  Warren|  null| paid|Spokane-Spokane V...|   GET|Cancellation Conf...|1532482662000|     1819|null|   200|1542051608000|Mozilla/5.0 (Wind...|    54|    1|           1|\n",
      "|  null|Cancelled|  Rodrigo|     M|           73|   Carey|  null| paid|New York-Newark-J...|   GET|Cancellation Conf...|1535389443000|      185|null|   200|1542740649000|\"Mozilla/5.0 (Win...|100014|    1|           1|\n",
      "|  null|Cancelled|     Alex|     M|          145|   Hogan|  null| paid|Denver-Aurora-Lak...|   GET|Cancellation Conf...|1535066380000|      842|null|   200|1539729037000|Mozilla/5.0 (Wind...|   101|    1|           1|\n",
      "|  null|Cancelled|   Rayden|     M|          185|Phillips|  null| free|Dallas-Fort Worth...|   GET|Cancellation Conf...|1537550344000|      172|null|   200|1543143782000|\"Mozilla/5.0 (Mac...|100021|    1|           1|\n",
      "|  null|Cancelled|    Kaleb|     M|           49|Thompson|  null| paid|Los Angeles-Long ...|   GET|Cancellation Conf...|1536988041000|     1864|null|   200|1542181031000|\"Mozilla/5.0 (Mac...|    29|    1|           1|\n",
      "|  null|Cancelled|   Rachel|     F|           11|  Bailey|  null| paid|Albany-Schenectad...|   GET|Cancellation Conf...|1536102943000|     1059|null|   200|1540402387000|Mozilla/5.0 (Wind...|    73|    1|           1|\n",
      "|  null|Cancelled|Christian|     M|          100|Robinson|  null| paid|       Quincy, IL-MO|   GET|Cancellation Conf...|1534942082000|     1092|null|   200|1540050556000|\"Mozilla/5.0 (Win...|    87|    1|           1|\n",
      "|  null|Cancelled|  Jeffery|     M|           46| Wheeler|  null| paid|         Bozeman, MT|   GET|Cancellation Conf...|1533886191000|     1324|null|   200|1540875543000|\"Mozilla/5.0 (Win...|     3|    1|           1|\n",
      "|  null|Cancelled|   Teagan|     F|          306| Roberts|  null| paid|New Philadelphia-...|   GET|Cancellation Conf...|1537634865000|     1519|null|   200|1541463632000|Mozilla/5.0 (Wind...|    28|    1|           1|\n",
      "|  null|Cancelled|     Mahi|     F|           17|Robinson|  null| free|Houston-The Woodl...|   GET|Cancellation Conf...|1537898335000|      191|null|   200|1542038645000|Mozilla/5.0 (Maci...|100022|    1|           1|\n",
      "|  null|Cancelled|   London|     F|           49|    Diaz|  null| free|New York-Newark-J...|   GET|Cancellation Conf...|1532625569000|      129|null|   200|1540299152000|\"Mozilla/5.0 (Mac...|100025|    1|           1|\n",
      "|  null|Cancelled|    Lucas|     M|          142|Franklin|  null| paid|Phoenix-Mesa-Scot...|   GET|Cancellation Conf...|1537707098000|       44|null|   200|1538705338000|\"Mozilla/5.0 (X11...|300007|    1|           1|\n",
      "|  null|Cancelled|    Leyla|     F|           45|  Barnes|  null| free|           Flint, MI|   GET|Cancellation Conf...|1537964483000|        6|null|   200|1538753070000|\"Mozilla/5.0 (Win...|100006|    1|           1|\n",
      "|  null|Cancelled|   Adriel|     M|          104| Mendoza|  null| paid|  Kansas City, MO-KS|   GET|Cancellation Conf...|1535623466000|      514|null|   200|1538943990000|\"Mozilla/5.0 (Mac...|    18|    1|           1|\n",
      "|  null|Cancelled|   Chayse|     M|           91|  Keller|  null| paid|         Jackson, MS|   GET|Cancellation Conf...|1529643103000|     1940|null|   200|1542234867000|\"Mozilla/5.0 (Win...|    70|    1|           1|\n",
      "|  null|Cancelled|    Cason|     M|           73|   Smith|  null| free|          Monroe, LA|   GET|Cancellation Conf...|1532610926000|      115|null|   200|1539971825000|\"Mozilla/5.0 (Mac...|100005|    1|           1|\n",
      "|  null|Cancelled|    Kayla|     F|          273| Johnson|  null| paid|Philadelphia-Camd...|   GET|Cancellation Conf...|1538333829000|      797|null|   200|1539465584000|Mozilla/5.0 (Wind...|    17|    1|           1|\n",
      "|  null|Cancelled|    Piper|     F|          201|    Cook|  null| paid|Little Rock-North...|   GET|Cancellation Conf...|1533522419000|      205|null|   200|1543491909000|\"Mozilla/5.0 (Win...|100007|    1|           1|\n",
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+--------------------+-------------+---------+----+------+-------------+--------------------+------+-----+------------+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.filter(F.col(\"Churn\") == 1).sort(\"num_of_churn\", ascending=False).show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that the user only cancel once, not subcribe and cancel again."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**There are some questions we should answer to understand the data:**\n",
    "- Difference between \"free\" and \"paid\" users in churn rate?\n",
    "- What difference between \"F\" and \"M\" users in churn rate?\n",
    "- What time in a day/month does the user cancel the subscription highest?\n",
    "- Does a user who is not very active result in cancellation?\n",
    "- Does the user still use Sparkify after canceling the subscription?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Difference between \"free\" and \"paid\" user in churn rate?**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_level = df.select(\"userId\", 'level', 'Churn').distinct()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f06e8c63be0>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_pd_level = df_level.select('level').toPandas()\n",
    "ax = sns.countplot(x='level', data=df_pd_level)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f06e8bf7e80>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_pd_churn = df_level.select('level', \"Churn\").toPandas()\n",
    "ax = sns.countplot(x='Churn', hue='level', data=df_pd_churn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that `paid` user has higher churn rate than `free` user."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Difference between \"F\" and \"M\" user in churn rate?**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_gender = df.select(\"userId\", 'gender', 'Churn').distinct()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f06bf10ce48>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_pd_gender = df_gender.select('gender').toPandas()\n",
    "ax = sns.countplot(x='gender', data=df_pd_gender)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f06bf0d7e48>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_pd_churn = df_gender.select('gender', 'Churn').toPandas()\n",
    "ax = sns.countplot(x='Churn', hue='gender', data=df_pd_churn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In general, the number of male user is greater than female and this trend remain the same for churn rate."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**What time in a month/year does user cancel the subcription highest?**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+-------------------+-----+----+-----+---+----+\n",
      "|userId|                 ts|Churn|year|month|day|hour|\n",
      "+------+-------------------+-----+----+-----+---+----+\n",
      "|    30|2018-10-01 00:01:57|    0|2018|   10|  1|   0|\n",
      "|     9|2018-10-01 00:03:00|    0|2018|   10|  1|   0|\n",
      "|    30|2018-10-01 00:06:34|    0|2018|   10|  1|   0|\n",
      "|     9|2018-10-01 00:06:56|    0|2018|   10|  1|   0|\n",
      "|    30|2018-10-01 00:11:16|    0|2018|   10|  1|   0|\n",
      "+------+-------------------+-----+----+-----+---+----+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_month = df.select('userId', 'ts', 'Churn')\n",
    "\n",
    "df_month = df_month.withColumn('ts', F.col('ts')/1000)\n",
    "\n",
    "df_month = df_month.withColumn('ts', F.to_timestamp(F.col('ts')))\n",
    "\n",
    "df_month = df_month.withColumn('year', F.year('ts')) \\\n",
    "    .withColumn('month', F.month('ts')) \\\n",
    "    .withColumn('day', F.dayofmonth('ts')) \\\n",
    "    .withColumn('hour', F.hour('ts'))\n",
    "\n",
    "df_month.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+----------+\n",
      "|month|sum(Churn)|\n",
      "+-----+----------+\n",
      "|   10|        30|\n",
      "|   11|        22|\n",
      "|   12|         0|\n",
      "+-----+----------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_month.groupBy('month').sum('Churn').sort('month').show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f06bf28a550>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_pd_month = df_month.filter(F.col('Churn') == 1).toPandas()\n",
    "plt.figure(figsize = (10, 5))\n",
    "ax = sns.countplot(x='day', hue='Churn', data=df_pd_month)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The day with the highest churn rate is at the beginning of the month, which is the 2nd."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Does user who not very active result in cancellation?**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------------+---------+---------+------+-------------+--------+---------+-----+------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+------------+\n",
      "|              artist|     auth|firstName|gender|itemInSession|lastName|   length|level|          location|method|    page| registration|sessionId|                song|status|           ts|           userAgent|userId|Churn|num_of_churn|\n",
      "+--------------------+---------+---------+------+-------------+--------+---------+-----+------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+------------+\n",
      "| Christopher O'Riley|Logged In|    Mason|     M|            0|    Hart|337.91955| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|    paranoid android|   200|1539317144000|\"Mozilla/5.0 (Mac...|   125|    0|           0|\n",
      "|The Notorious B.I.G.|Logged In|    Mason|     M|            1|    Hart|230.03383| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|Hypnotize(Album V...|   200|1539317481000|\"Mozilla/5.0 (Mac...|   125|    0|           0|\n",
      "|           Betty Boo|Logged In|    Mason|     M|            2|    Hart| 203.2322| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|       I'm On My Way|   200|1539317711000|\"Mozilla/5.0 (Mac...|   125|    0|           0|\n",
      "|          Nickelback|Logged In|    Mason|     M|            3|    Hart|210.83383| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|Leader Of Men (Al...|   200|1539317914000|\"Mozilla/5.0 (Mac...|   125|    0|           0|\n",
      "| Ready For The World|Logged In|    Mason|     M|            4|    Hart|391.26159| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|       Love You Down|   200|1539318124000|\"Mozilla/5.0 (Mac...|   125|    0|           0|\n",
      "+--------------------+---------+---------+------+-------------+--------+---------+-----+------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+------------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_churn_user = df.filter(F.col('Churn') == 1)\n",
    "\n",
    "df_churn_user = df_churn_user.select(F.col('userId').alias('churn_userId'))\n",
    "\n",
    "df_churn_log = df.join(\n",
    "    df_churn_user,\n",
    "    df.userId == df_churn_user.churn_userId,\n",
    "    'inner'\n",
    ").drop('churn_userId')\n",
    "\n",
    "df_churn_log.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "44864"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_churn_log.count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f06bc6f9208>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_pd_dist = df_churn_log.groupBy('userId').count().toPandas()\n",
    "ax = sns.distplot(df_pd_dist['count'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The user is not very active before canceling the subscription."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Does user still use Sparkify after cancel the subscription?**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------------+---------+---------+------+-------------+--------+---------+-----+------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+------------+-------------+\n",
      "|              artist|     auth|firstName|gender|itemInSession|lastName|   length|level|          location|method|    page| registration|sessionId|                song|status|           ts|           userAgent|userId|Churn|num_of_churn|       max_ts|\n",
      "+--------------------+---------+---------+------+-------------+--------+---------+-----+------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+------------+-------------+\n",
      "| Christopher O'Riley|Logged In|    Mason|     M|            0|    Hart|337.91955| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|    paranoid android|   200|1539317144000|\"Mozilla/5.0 (Mac...|   125|    0|           0|1539318918000|\n",
      "|The Notorious B.I.G.|Logged In|    Mason|     M|            1|    Hart|230.03383| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|Hypnotize(Album V...|   200|1539317481000|\"Mozilla/5.0 (Mac...|   125|    0|           0|1539318918000|\n",
      "|           Betty Boo|Logged In|    Mason|     M|            2|    Hart| 203.2322| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|       I'm On My Way|   200|1539317711000|\"Mozilla/5.0 (Mac...|   125|    0|           0|1539318918000|\n",
      "|          Nickelback|Logged In|    Mason|     M|            3|    Hart|210.83383| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|Leader Of Men (Al...|   200|1539317914000|\"Mozilla/5.0 (Mac...|   125|    0|           0|1539318918000|\n",
      "| Ready For The World|Logged In|    Mason|     M|            4|    Hart|391.26159| free|Corpus Christi, TX|   PUT|NextSong|1533157139000|      174|       Love You Down|   200|1539318124000|\"Mozilla/5.0 (Mac...|   125|    0|           0|1539318918000|\n",
      "+--------------------+---------+---------+------+-------------+--------+---------+-----+------------------+------+--------+-------------+---------+--------------------+------+-------------+--------------------+------+-----+------------+-------------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "window = Window.partitionBy(\"userId\").orderBy(\"ts\").rowsBetween(Window.unboundedPreceding, Window.unboundedFollowing)\n",
    "df_cancel = df_churn_log.withColumn(\"max_ts\", F.max(\"ts\").over(window))\n",
    "df_cancel.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+--------------------+-------------+---------+----+------+-------------+--------------------+------+-----+------------+-------------+\n",
      "|artist|     auth|firstName|gender|itemInSession|lastName|length|level|            location|method|                page| registration|sessionId|song|status|           ts|           userAgent|userId|Churn|num_of_churn|       max_ts|\n",
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+--------------------+-------------+---------+----+------+-------------+--------------------+------+-----+------------+-------------+\n",
      "|  null|Logged In|    Mason|     M|            9|    Hart|  null| free|  Corpus Christi, TX|   PUT|              Cancel|1533157139000|      174|null|   307|1539318918000|\"Mozilla/5.0 (Mac...|   125|    0|           0|1539318918000|\n",
      "|  null|Cancelled|    Mason|     M|           10|    Hart|  null| free|  Corpus Christi, TX|   GET|Cancellation Conf...|1533157139000|      174|null|   200|1539318918000|\"Mozilla/5.0 (Mac...|   125|    1|           1|1539318918000|\n",
      "|  null|Cancelled|    Ethan|     M|          176| Johnson|  null| paid|Lexington-Fayette...|   GET|Cancellation Conf...|1538080987000|      934|null|   200|1539761972000|\"Mozilla/5.0 (Win...|    51|    1|           1|1539761972000|\n",
      "|  null|Cancelled|    Alexi|     F|           42|  Warren|  null| paid|Spokane-Spokane V...|   GET|Cancellation Conf...|1532482662000|     1819|null|   200|1542051608000|Mozilla/5.0 (Wind...|    54|    1|           1|1542051608000|\n",
      "|  null|Cancelled|  Rodrigo|     M|           73|   Carey|  null| paid|New York-Newark-J...|   GET|Cancellation Conf...|1535389443000|      185|null|   200|1542740649000|\"Mozilla/5.0 (Win...|100014|    1|           1|1542740649000|\n",
      "|  null|Cancelled|     Alex|     M|          145|   Hogan|  null| paid|Denver-Aurora-Lak...|   GET|Cancellation Conf...|1535066380000|      842|null|   200|1539729037000|Mozilla/5.0 (Wind...|   101|    1|           1|1539729037000|\n",
      "|  null|Cancelled|   Rayden|     M|          185|Phillips|  null| free|Dallas-Fort Worth...|   GET|Cancellation Conf...|1537550344000|      172|null|   200|1543143782000|\"Mozilla/5.0 (Mac...|100021|    1|           1|1543143782000|\n",
      "|  null|Cancelled|    Kaleb|     M|           49|Thompson|  null| paid|Los Angeles-Long ...|   GET|Cancellation Conf...|1536988041000|     1864|null|   200|1542181031000|\"Mozilla/5.0 (Mac...|    29|    1|           1|1542181031000|\n",
      "|  null|Cancelled|   Rachel|     F|           11|  Bailey|  null| paid|Albany-Schenectad...|   GET|Cancellation Conf...|1536102943000|     1059|null|   200|1540402387000|Mozilla/5.0 (Wind...|    73|    1|           1|1540402387000|\n",
      "|  null|Cancelled|Christian|     M|          100|Robinson|  null| paid|       Quincy, IL-MO|   GET|Cancellation Conf...|1534942082000|     1092|null|   200|1540050556000|\"Mozilla/5.0 (Win...|    87|    1|           1|1540050556000|\n",
      "|  null|Cancelled|  Jeffery|     M|           46| Wheeler|  null| paid|         Bozeman, MT|   GET|Cancellation Conf...|1533886191000|     1324|null|   200|1540875543000|\"Mozilla/5.0 (Win...|     3|    1|           1|1540875543000|\n",
      "|  null|Cancelled|   Teagan|     F|          306| Roberts|  null| paid|New Philadelphia-...|   GET|Cancellation Conf...|1537634865000|     1519|null|   200|1541463632000|Mozilla/5.0 (Wind...|    28|    1|           1|1541463632000|\n",
      "|  null|Cancelled|     Mahi|     F|           17|Robinson|  null| free|Houston-The Woodl...|   GET|Cancellation Conf...|1537898335000|      191|null|   200|1542038645000|Mozilla/5.0 (Maci...|100022|    1|           1|1542038645000|\n",
      "|  null|Cancelled|   London|     F|           49|    Diaz|  null| free|New York-Newark-J...|   GET|Cancellation Conf...|1532625569000|      129|null|   200|1540299152000|\"Mozilla/5.0 (Mac...|100025|    1|           1|1540299152000|\n",
      "|  null|Cancelled|    Lucas|     M|          142|Franklin|  null| paid|Phoenix-Mesa-Scot...|   GET|Cancellation Conf...|1537707098000|       44|null|   200|1538705338000|\"Mozilla/5.0 (X11...|300007|    1|           1|1538705338000|\n",
      "|  null|Cancelled|    Leyla|     F|           45|  Barnes|  null| free|           Flint, MI|   GET|Cancellation Conf...|1537964483000|        6|null|   200|1538753070000|\"Mozilla/5.0 (Win...|100006|    1|           1|1538753070000|\n",
      "|  null|Cancelled|   Adriel|     M|          104| Mendoza|  null| paid|  Kansas City, MO-KS|   GET|Cancellation Conf...|1535623466000|      514|null|   200|1538943990000|\"Mozilla/5.0 (Mac...|    18|    1|           1|1538943990000|\n",
      "|  null|Cancelled|   Chayse|     M|           91|  Keller|  null| paid|         Jackson, MS|   GET|Cancellation Conf...|1529643103000|     1940|null|   200|1542234867000|\"Mozilla/5.0 (Win...|    70|    1|           1|1542234867000|\n",
      "|  null|Cancelled|    Cason|     M|           73|   Smith|  null| free|          Monroe, LA|   GET|Cancellation Conf...|1532610926000|      115|null|   200|1539971825000|\"Mozilla/5.0 (Mac...|100005|    1|           1|1539971825000|\n",
      "|  null|Cancelled|    Kayla|     F|          273| Johnson|  null| paid|Philadelphia-Camd...|   GET|Cancellation Conf...|1538333829000|      797|null|   200|1539465584000|Mozilla/5.0 (Wind...|    17|    1|           1|1539465584000|\n",
      "+------+---------+---------+------+-------------+--------+------+-----+--------------------+------+--------------------+-------------+---------+----+------+-------------+--------------------+------+-----+------------+-------------+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_cancel = df_cancel.filter(F.col('ts') == F.col('max_ts'))\n",
    "df_cancel.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Row(artist=None)]"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_cancel.select('artist').distinct().collect()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After canceling the subscription, no one continues to use Sparkify"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature Engineering\n",
    "Once you've familiarized yourself with the data, build out the features you find promising to train your model on. To work with the full dataset, you can follow the following steps.\n",
    "- Write a script to extract the necessary features from the smaller subset of data\n",
    "- Ensure that your script is scalable, using the best practices discussed in Lesson 3\n",
    "- Try your script on the full data set, debugging your script if necessary\n",
    "\n",
    "If you are working in the classroom workspace, you can just extract features based on the small subset of data contained here. Be sure to transfer over this work to the larger dataset when you work on your Spark cluster."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Feature for the model:**\n",
    "- `Churn`\n",
    "- `level`\n",
    "- `gender`\n",
    "- `avg_song_length`\n",
    "- `total_artist`\n",
    "- `total_song`\n",
    "- `total_session`\n",
    "- `total_thumb_up`\n",
    "- `total_thumb_down`\n",
    "- `total_add_friend`\n",
    "- `total_add_playlist`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_feature = df.select('userId', 'level', 'gender', 'Churn', 'ts', 'length', 'artist', 'song', 'page')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+\n",
      "|userId|   avg_song_length|total_artist|total_song|total_session|total_thumb_up|total_thumb_down|total_add_friend|total_add_playlist|\n",
      "+------+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+\n",
      "|100010|  243.421444909091|         252|       269|          381|            17|               5|               4|                 7|\n",
      "|200002|242.91699209302305|         339|       378|          474|            21|               6|               4|                 8|\n",
      "|   125|261.13913750000006|           8|         8|           11|             0|               0|               0|                 0|\n",
      "|    51|247.88055082899118|        1385|      1854|         2464|           100|              21|              28|                52|\n",
      "|   124|248.17653659965674|        2232|      3339|         4825|           171|              41|              74|               118|\n",
      "+------+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Create features\n",
    "\n",
    "df_by_user = df_feature.groupBy('userId').agg(\n",
    "    F.avg('length').alias('avg_song_length'),\n",
    "    F.countDistinct('artist').alias('total_artist'),\n",
    "    F.countDistinct('song').alias('total_song'),\n",
    "    F.count('page').alias('total_session'),\n",
    "    F.count(F.when(F.col('page') == 'Thumbs Up', True)).alias('total_thumb_up'),\n",
    "    F.count(F.when(F.col('page') == 'Thumbs Down', True)).alias('total_thumb_down'),\n",
    "    F.count(F.when(F.col('page') == 'Add Friend', True)).alias('total_add_friend'),\n",
    "    F.count(F.when(F.col('page') == 'Add to Playlist', True)).alias('total_add_playlist')\n",
    ")\n",
    "df_by_user.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+-----+------+-----+-------------+------+------+----+--------------------+-------------+\n",
      "|userId|level|gender|Churn|           ts|length|artist|song|                page|       max_ts|\n",
      "+------+-----+------+-----+-------------+------+------+----+--------------------+-------------+\n",
      "|100010| free|     F|    0|1542823952000|  null|  null|null|              Logout|1542823952000|\n",
      "|200002| paid|     M|    0|1542324354000|  null|  null|null|           Thumbs Up|1542324354000|\n",
      "|   125| free|     M|    0|1539318918000|  null|  null|null|              Cancel|1539318918000|\n",
      "|   125| free|     M|    1|1539318918000|  null|  null|null|Cancellation Conf...|1539318918000|\n",
      "|   124| paid|     F|    0|1543590766000|  null|  null|null|                Home|1543590766000|\n",
      "+------+-----+------+-----+-------------+------+------+----+--------------------+-------------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Keep latest records\n",
    "\n",
    "window = Window.partitionBy(\"userId\").orderBy(\"ts\").rowsBetween(Window.unboundedPreceding, Window.unboundedFollowing)\n",
    "df_user = df_feature.withColumn(\"max_ts\", F.max(\"ts\").over(window))\n",
    "df_user = df_user.filter(F.col('ts') == F.col('max_ts'))\n",
    "df_user.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+-----+------+-----+-------------+------+------+----+--------------------+-------------+---------+\n",
      "|userId|level|gender|Churn|           ts|length|artist|song|                page|       max_ts|sum_churn|\n",
      "+------+-----+------+-----+-------------+------+------+----+--------------------+-------------+---------+\n",
      "|100010| free|     F|    0|1542823952000|  null|  null|null|              Logout|1542823952000|        0|\n",
      "|200002| paid|     M|    0|1542324354000|  null|  null|null|           Thumbs Up|1542324354000|        0|\n",
      "|   125| free|     M|    1|1539318918000|  null|  null|null|Cancellation Conf...|1539318918000|        1|\n",
      "|   124| paid|     F|    0|1543590766000|  null|  null|null|                Home|1543590766000|        0|\n",
      "|    51| paid|     M|    1|1539761972000|  null|  null|null|Cancellation Conf...|1539761972000|        1|\n",
      "+------+-----+------+-----+-------------+------+------+----+--------------------+-------------+---------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Keep record that \"Churn\" = 1, otherwise 0\n",
    "\n",
    "df_user = df_user.withColumn('sum_churn', F.sum('Churn').over(Window.partitionBy(\"userId\")))\n",
    "df_user = df_user.filter(~((F.col('Churn') == 0) & (F.col('sum_churn') > 0)))\n",
    "\n",
    "df_user.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+-----+------+-----+\n",
      "|   avg_song_length|total_artist|total_song|total_session|total_thumb_up|total_thumb_down|total_add_friend|total_add_playlist|level|gender|Churn|\n",
      "+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+-----+------+-----+\n",
      "|  243.421444909091|         252|       269|          381|            17|               5|               4|                 7| free|     F|    0|\n",
      "|242.91699209302305|         339|       378|          474|            21|               6|               4|                 8| paid|     M|    0|\n",
      "|261.13913750000006|           8|         8|           11|             0|               0|               0|                 0| free|     M|    1|\n",
      "|248.17653659965674|        2232|      3339|         4825|           171|              41|              74|               118| paid|     F|    0|\n",
      "|247.88055082899118|        1385|      1854|         2464|           100|              21|              28|                52| paid|     M|    1|\n",
      "+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+-----+------+-----+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_user = df_user.select('userId', 'level', 'gender', 'Churn').distinct()\n",
    "\n",
    "df_merge = df_by_user.join(df_user, 'userId', 'left').drop('userId')\n",
    "df_merge.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+-----+------+-----+\n",
      "|   avg_song_length|total_artist|total_song|total_session|total_thumb_up|total_thumb_down|total_add_friend|total_add_playlist|level|gender|Churn|\n",
      "+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+-----+------+-----+\n",
      "|  243.421444909091|         252|       269|          381|            17|               5|               4|                 7|    0|     0|    0|\n",
      "|242.91699209302305|         339|       378|          474|            21|               6|               4|                 8|    1|     1|    0|\n",
      "|261.13913750000006|           8|         8|           11|             0|               0|               0|                 0|    0|     1|    1|\n",
      "|248.17653659965674|        2232|      3339|         4825|           171|              41|              74|               118|    1|     0|    0|\n",
      "|247.88055082899118|        1385|      1854|         2464|           100|              21|              28|                52|    1|     1|    1|\n",
      "+------------------+------------+----------+-------------+--------------+----------------+----------------+------------------+-----+------+-----+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df_final = df_merge.withColumn(\"level\", F.when(F.col(\"level\") == \"paid\", 1).otherwise(0)) \\\n",
    "    .withColumn(\"gender\", F.when(F.col(\"gender\") == \"M\", 1).otherwise(0))\n",
    "\n",
    "df_final.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.6/site-packages/IPython/core/interactiveshell.py:2961: UserWarning: To output multiple subplots, the figure containing the passed axes is being cleared\n",
      "  exec(code_obj, self.user_global_ns, self.user_ns)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f06bc663470>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_pd_final = df_final.toPandas()\n",
    "\n",
    "fig = plt.figure(figsize=(30, 25))\n",
    "ax = fig.gca()\n",
    "h = df_pd_final.hist(ax=ax)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Modeling\n",
    "Split the full dataset into train, test, and validation sets. Test out several of the machine learning methods I learned. Evaluate the accuracy of the various models, tuning parameters as necessary. Determine my winning model based on test accuracy and report results on the validation set. Since the churned users are a fairly small subset, I used F1 score as the metric to optimize."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_columns = df_final.columns[:-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "train, test = df_final.randomSplit([0.7, 0.3], seed=12)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_model(classifier, param):\n",
    "    \"\"\"\n",
    "    Build the classfication model with Cross Validation\n",
    "    \n",
    "    Parameters:\n",
    "    classifier: Model\n",
    "    param: parameter tuning for the model\n",
    "    \n",
    "    Returns:\n",
    "    Classfication model\n",
    "    \"\"\"\n",
    "    \n",
    "    assembler = VectorAssembler(inputCols=feature_columns, outputCol=\"features\")\n",
    "    scaler = MinMaxScaler(inputCol=\"features\", outputCol=\"scaled_features\")\n",
    "    pipeline = Pipeline(stages=[assembler, scaler, classifier])\n",
    "\n",
    "    model = CrossValidator(\n",
    "        estimator=pipeline,\n",
    "        estimatorParamMaps=param,\n",
    "        evaluator=MulticlassClassificationEvaluator(labelCol='Churn', metricName='f1'),\n",
    "        numFolds=5,\n",
    "    )\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate_model(pred):\n",
    "    \"\"\" Model evaluation with F1 score metric \"\"\"\n",
    "    \n",
    "    evaluator = MulticlassClassificationEvaluator(predictionCol=\"prediction\", labelCol=\"Churn\")\n",
    "    f1_score = evaluator.evaluate(pred, {evaluator.metricName: \"f1\"})\n",
    "    print('F1 score: {}'.format(f1_score))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = LogisticRegression(featuresCol=\"scaled_features\", labelCol=\"Churn\")\n",
    "param = ParamGridBuilder().build()\n",
    "model = build_model(lr, param)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "fit_model = model.fit(train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred = fit_model.transform(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 score: 0.6744862155388471\n"
     ]
    }
   ],
   "source": [
    "evaluate_model(pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "rf = RandomForestClassifier(featuresCol=\"scaled_features\", labelCol=\"Churn\")\n",
    "rf_param = ParamGridBuilder().build()\n",
    "rf_model = build_model(rf, rf_param)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train time: 378.3478693962097\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "rf_fit_model = rf_model.fit(train)\n",
    "print(\"Train time:\", time.time() - start_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "rf_pred = rf_fit_model.transform(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 score: 0.6474418604651163\n"
     ]
    }
   ],
   "source": [
    "evaluate_model(rf_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Gradient Boosting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "gbt = GBTClassifier(featuresCol=\"scaled_features\", labelCol=\"Churn\")\n",
    "gbt_param = ParamGridBuilder().build()\n",
    "gbt_model = build_model(gbt, gbt_param)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train time: 1719.4316709041595\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "gbt_fit_model = gbt_model.fit(train)\n",
    "print(\"Train time:\", time.time() - start_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "gbt_pred = gbt_fit_model.transform(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 score: 0.6883597883597884\n"
     ]
    }
   ],
   "source": [
    "evaluate_model(gbt_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Try different settings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "gbt = GBTClassifier(featuresCol=\"scaled_features\", labelCol=\"Churn\")\n",
    "gbt_param = ParamGridBuilder() \\\n",
    "    .addGrid(gbt.maxDepth, [3, 5, 7]) \\\n",
    "    .build()\n",
    "gbt_model = build_model(gbt, gbt_param)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train time: 4439.241886377335\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "gbt_fit_model = gbt_model.fit(train)\n",
    "print(\"Train time:\", time.time() - start_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "gbt_pred = gbt_fit_model.transform(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 score: 0.6571957671957671\n"
     ]
    }
   ],
   "source": [
    "evaluate_model(gbt_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameter for maxDepth: 3\n"
     ]
    }
   ],
   "source": [
    "bestPipeline = gbt_fit_model.bestModel\n",
    "print(\"Best parameter for maxDepth:\", bestPipeline.stages[2].getOrDefault('maxDepth'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "avg_song_length : 0.08716097168631914 \n",
      "\n",
      "total_artist : 0.14142548996512083 \n",
      "\n",
      "total_song : 0.07343415697469002 \n",
      "\n",
      "total_session : 0.06644850369535893 \n",
      "\n",
      "total_thumb_up : 0.13608297448699597 \n",
      "\n",
      "total_thumb_down : 0.1720680267372067 \n",
      "\n",
      "total_add_friend : 0.25710134523964884 \n",
      "\n",
      "total_add_playlist : 0.01411748393361805 \n",
      "\n",
      "level : 0.02460178536241849 \n",
      "\n",
      "gender : 0.02755926191862303 \n",
      "\n"
     ]
    }
   ],
   "source": [
    "for i in range(len(bestPipeline.stages[2].featureImportances)):\n",
    "    print(\"{} : {} \\n\".format(df_final.columns[i], bestPipeline.stages[2].featureImportances[i]))"
   ]
  },
  {
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}