Spark One Hot Encoding.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from pyspark import SparkConf\n",
    "from pyspark.sql import SparkSession\n",
    "from pyspark.ml.feature import StringIndexer\n",
    "from pyspark.ml.feature import OneHotEncoder\n",
    "from pyspark.ml import Pipeline\n",
    "from pyspark.sql.functions import when\n",
    "\n",
    "sparkConf = SparkConf(loadDefaults=True).setAppName(\"Encoding\")\n",
    "spark = SparkSession.builder.master(\"local\").config(conf=sparkConf).getOrCreate()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Load the data as a Spark Dataframe now and select the same columns for One Hot Encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------------+--------------+--------------------+\n",
      "|        GenderSelect|       Country|    EmploymentStatus|\n",
      "+--------------------+--------------+--------------------+\n",
      "|Non-binary, gende...|          null|  Employed full-time|\n",
      "|              Female| United States|Not employed, but...|\n",
      "|                Male|        Canada|Not employed, but...|\n",
      "|                Male| United States|Independent contr...|\n",
      "|                Male|        Taiwan|  Employed full-time|\n",
      "|                Male|        Brazil|  Employed full-time|\n",
      "|                Male| United States|  Employed full-time|\n",
      "|              Female|         India|  Employed full-time|\n",
      "|              Female|     Australia|  Employed full-time|\n",
      "|                Male|        Russia|  Employed full-time|\n",
      "|              Female|        Russia|Not employed, and...|\n",
      "|                Male|         India|  Employed full-time|\n",
      "|                Male|        Brazil|  Employed full-time|\n",
      "|                Male|   Netherlands|  Employed full-time|\n",
      "|                Male|        Taiwan|  Employed full-time|\n",
      "|                Male| United States|Independent contr...|\n",
      "|                Male|         Italy|  Employed full-time|\n",
      "|                Male|United Kingdom|  Employed full-time|\n",
      "|                Male| United States|Not employed, but...|\n",
      "|                Male|        Brazil|Not employed, but...|\n",
      "+--------------------+--------------+--------------------+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "sparkDF = spark.read.csv('multipleChoiceResponses.csv',header=True)\n",
    "sparkDF = sparkDF.limit(100)\n",
    "catColumns = ['GenderSelect','Country','EmploymentStatus']\n",
    "sparkDF.select(catColumns).show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Need to reset empty cell values to __NA__ otherwise subsequent operation of StringIndexer fails"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "for col in catColumns:\n",
    "    sparkDF = sparkDF.withColumn(col, when(sparkDF[col].isNull(),\"__NA__\").otherwise(sparkDF[col]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Create a pipeline that first converts categorical values to numerical using StringIndexer and then applies OneHotEncoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "indexers = [StringIndexer(inputCol=col, outputCol=col+\"_encoded\").fit(sparkDF) for col in catColumns]\n",
    "encoders = [OneHotEncoder(inputCol=col+\"_encoded\", outputCol=col+\"_onehot\") for col in catColumns]\n",
    "p1 = Pipeline(stages=indexers)\n",
    "indexedDF = p1.fit(sparkDF).transform(sparkDF)\n",
    "p2 = Pipeline(stages=encoders)\n",
    "encodedDF = p2.fit(indexedDF).transform(indexedDF)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Show the dense vector representation of the one hot encoded vectors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-------------------+---------------+-----------------------+\n",
      "|GenderSelect_onehot| Country_onehot|EmploymentStatus_onehot|\n",
      "+-------------------+---------------+-----------------------+\n",
      "|          (2,[],[])|(31,[14],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[1],[1.0])| (31,[0],[1.0])|          (5,[1],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[3],[1.0])|          (5,[1],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[0],[1.0])|          (5,[3],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[8],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[7],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[0],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[1],[1.0])| (31,[1],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[1],[1.0])|(31,[27],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[2],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[1],[1.0])| (31,[2],[1.0])|          (5,[2],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[1],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[7],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])|(31,[17],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[8],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[0],[1.0])|          (5,[3],[1.0])|\n",
      "|      (2,[0],[1.0])|(31,[28],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[6],[1.0])|          (5,[0],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[0],[1.0])|          (5,[1],[1.0])|\n",
      "|      (2,[0],[1.0])| (31,[7],[1.0])|          (5,[1],[1.0])|\n",
      "+-------------------+---------------+-----------------------+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "encodedDF.select('GenderSelect_onehot','Country_onehot','EmploymentStatus_onehot').show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}