Introduction to PySpark

Introduction to PySpark

• 1.
  A One HourIntroduction to Analytics with PySpark Introduction to PySpark http://www.slideshare.net/rjurney/introduction-to-pyspark or http://bit.ly/intro_to_pyspark
• 2.
  Agile Data Science2.0 Russell Jurney 2 Data Engineer Data Scientist Visualization Software Engineer 85% 85% 85% Writer 85% Teacher 50% Russell Jurney is a veteran data scientist and thought leader. He coined the term Agile Data Science in the book of that name from O’Reilly in 2012, which outlines the first agile development methodology for data science. Russell has constructed numerous full-stack analytics products over the past ten years and now works with clients helping them extract value from their data assets. Russell Jurney Skill Principal Consultant at Data Syndrome Russell Jurney Data Syndrome, LLC Email : [email protected] Web : datasyndrome.com Principal Consultant
• 3.
  Building Full-Stack DataAnalytics Applications with Spark http://bit.ly/agile_data_science Agile Data Science 2.0
• 4.
  Agile Data Science2.0 4 Realtime Predictive Analytics Rapidly learn to build entire predictive systems driven by Kafka, PySpark, Speak Streaming, Spark MLlib and with a web front-end using Python/Flask and JQuery. Available for purchase at http://datasyndrome.com/video
• 5.
  Lorem Ipsum dolorsiamet suame this placeholder for text can simply random text. It has roots in a piece of classical. variazioni deiwords which whichhtly. ven on your zuniga merida della is not denis. Product Consulting We build analytics products and systems consisting of big data viz, predictions, recommendations, reports and search. Corporate Training We offer training courses for data scientists and engineers and data science teams, Video Training We offer video training courses that rapidly acclimate you with a technology and technique.
• 6.
  Agile Data Science2.0 6 What is Spark? What makes it go? Concepts
• 7.
  Data Syndrome: AgileData Science 2.0 Hadoop / HDFS HDFS splits large data among many machines 7
• 8.
  Data Syndrome: AgileData Science 2.0 Hadoop / MapReduce In the beginning there was MapReduce 8
• 9.
  Data Syndrome: AgileData Science 2.0 Spark / RDD Spark RDDs are iterable MapReduce relations 9
• 10.
  Data Syndrome: AgileData Science 2.0 Spark / DataFrame Fast SQLish RDD thingies 10
• 11.
  Data Syndrome: AgileData Science 2.0 Spark Streaming Spark on realtime streams in mini-batches 11
• 12.
  Data Syndrome: AgileData Science 2.0 Spark Ecosystem Lots of cool stuff working together… 12 /
• 13.
  Agile Data Science2.0 13 Setting up our class environment Setup
• 14.
  Data Syndrome: AgileData Science 2.0 14 Python 3 > 2.7 While the break in compatibility between Python 2.X and 3.X was unfortunate and unnecessary , Python 3 has increasingly become the platform of choice for analytics work. With a few alterations, all code in this course will execute in a Python 2.7 environment.
• 15.
  Data Syndrome: AgileData Science 2.0 15 Virtualbox Virtualbox is a Free and Open Source Software (FOSS) virtualization product for AMD64/Intel64 processors. It supports many operating systems, and is under active development. https://www.virtualbox.org/wiki/Downloads
• 16.
  Data Syndrome: AgileData Science 2.0 16 Vagrant Vagrant sits on top of Virtualbox and provides easy to use, reproducible development environments. https://www.vagrantup.com/downloads.html
• 17.
  Data Syndrome: AgileData Science 2.0 Vagrant Setup 17 Initializing our Vagrant Environment # Get the project git clone https://github.com/rjurney/Agile_Data_Code_2/ # Setup and connect to our virtual machine vagrant up; vagrant ssh # Now, from within Vagrant cd Agile_Data_Code_2 intro_download.sh # See Appendix A and install.sh for manual install
• 18.
  Data Syndrome: AgileData Science 2.0 18 Amazon EC2 Alternatively, Amazon Web Services provide a simple way to launch a prepared image for use in this exercise.
• 19.
  Data Syndrome: AgileData Science 2.0 EC2 Setup for Ubuntu Linux 19 Initializing our EC2 Environment # See ec2.sh, which uses aws/ec2_bootstrap.sh # To use add: —user-data file://aws/ec2_bootstrap.sh # Get the project git clone [email protected]:rjurney/Agile_Data_Code_2.git # Setup AWS CLI tools pip install awscli # Edit and run r3.xlarge instance with your key ./ec2.sh # ssh to the machine
• 20.
  Data Syndrome: AgileData Science 2.0 EC2 Setup for Ubuntu Linux Initializing our EC2 Environment 20 # Contents of ec2.sh # Launch our instance, which ec2_bootstrap.sh will initialize
 aws ec2 run-instances 
 --image-id ami-4ae1fb5d 
 --key-name agile_data_science 
 --user-data file://aws/ec2_bootstrap.sh 
 --instance-type r3.xlarge 
 --ebs-optimized 
 --placement "AvailabilityZone=us-east-1d" 
 --block-device-mappings '{"DeviceName":"/dev/sda1","Ebs": {"DeleteOnTermination":false,"VolumeSize":1024}}' 
 --count 1

• 21.
  Data Syndrome: AgileData Science 2.0 EC2 Setup for Ubuntu Linux Initializing our EC2 Environment 21 # Download the data cd Agile_Data_Code_2 ./intro_download.sh
• 22.
  Data Syndrome: AgileData Science 2.0 EC2 Setup for Ubuntu Linux Initializing our EC2 Environment 22 # Download the data cd Agile_Data_Code_2 ./intro_download.sh
• 23.
  Data Syndrome: AgileData Science 2.0 Documentation Setup Opening the right web pages to answer your questions 23 http://spark.apache.org/docs/latest/api/python/pyspark.html http://spark.apache.org/docs/latest/api/python/pyspark.sql.html http://spark.apache.org/docs/latest/api/python/pyspark.ml.html
• 24.
  Agile Data Science2.0 24 Learning the basics of PySpark Basic PySpark
• 25.
  Data Syndrome: AgileData Science 2.0 Hello, World! How to load data and perform an operation on it in Spark 25 # See ch02/spark.py # Load the text file using the SparkContext
 csv_lines = sc.textFile("data/example.csv")
 
 # Map the data to split the lines into a list
 data = csv_lines.map(lambda line: line.split(","))
 
 # Collect the dataset into local RAM
 data.collect()
• 26.
  Data Syndrome: AgileData Science 2.0 Creating Objects from CSV using a function How to create objects from CSV using a function instead of a lambda 26 # See ch02/groupby.py csv_lines = sc.textFile("data/example.csv")
 
 # Turn the CSV lines into objects
 def csv_to_record(line):
 parts = line.split(",")
 record = {
 "name": parts[0],
 "company": parts[1],
 "title": parts[2]
 }
 return record
 
 # Apply the function to every record
 records = csv_lines.map(csv_to_record)
 
 # Inspect the first item in the dataset
 records.first()
• 27.
  Data Syndrome: AgileData Science 2.0 Using a GroupBy to Count Jobs Count things using the groupBy API 27 # Group the records by the name of the person
 grouped_records = records.groupBy(lambda x: x["name"])
 
 # Show the first group
 grouped_records.first()
 
 # Count the groups
 job_counts = grouped_records.map(
 lambda x: {
 "name": x[0],
 "job_count": len(x[1])
 }
 )
 
 job_counts.first()
 
 job_counts.collect()
• 28.
  Data Syndrome: AgileData Science 2.0 Map vs FlatMap Understanding the difference between these two operators 28 # See ch02/flatmap.py csv_lines = sc.textFile("data/example.csv")
 
 # Compute a relation of words by line
 words_by_line = csv_lines
 .map(lambda line: line.split(","))
 
 words_by_line.collect()
 
 # Compute a relation of words
 flattened_words = csv_lines
 .map(lambda line: line.split(","))
 .flatMap(lambda x: x)
 
 flattened_words.collect()
• 29.
  Data Syndrome: AgileData Science 2.0 Map vs FlatMap Understanding the difference between these two operators 29 words_by_line.collect() [['Russell Jurney', 'Relato', 'CEO'], ['Florian Liebert', 'Mesosphere', 'CEO'], ['Don Brown', 'Rocana', 'CIO'], ['Steve Jobs', 'Apple', 'CEO'], ['Donald Trump', 'The Trump Organization', 'CEO'], ['Russell Jurney', 'Data Syndrome', 'Principal Consultant']] flattened_words.collect() ['Russell Jurney', 'Relato', 'CEO', 'Florian Liebert', 'Mesosphere', 'CEO', 'Don Brown', 'Rocana', 'CIO', 'Steve Jobs', 'Apple', 'CEO', 'Donald Trump', 'The Trump Organization', 'CEO', 'Russell Jurney', 'Data Syndrome', 'Principal Consultant']
• 30.
  Data Syndrome: AgileData Science 2.0 Using DataFrames and Spark SQL to Count Jobs Converting an RDD to a DataFrame to use Spark SQL 30 # See ch02/sql.py csv_lines = sc.textFile("data/example.csv")
 
 from pyspark.sql import Row
 
 # Convert the CSV into a pyspark.sql.Row
 def csv_to_row(line):
 parts = line.split(",")
 row = Row(
 name=parts[0],
 company=parts[1],
 title=parts[2]
 )
 return row
 
 # Apply the function to get rows in an RDD
 rows = csv_lines.map(csv_to_row)
• 31.
  Data Syndrome: AgileData Science 2.0 Using DataFrames and Spark SQL to Count Jobs Converting an RDD to a DataFrame to use Spark SQL 31 # Convert to a pyspark.sql.DataFrame
 rows_df = rows.toDF()
 
 # Register the DataFrame for Spark SQL
 rows_df.registerTempTable("executives")
 
 # Generate a new DataFrame with SQL using the SparkSession
 job_counts = spark.sql(""" SELECT name, COUNT(*) AS total FROM executives GROUP BY name """)
 job_counts.show()
 
 # Go back to an RDD
 job_counts.rdd.collect()
• 32.
  Agile Data Science2.0 32 Working with a more complex dataset Exploratory Data Analysis with Airline Data
• 33.
  Data Syndrome: AgileData Science 2.0 Loading a Parquet Columnar File Using the Apache Parquet format to load columnar data 33 # See ch02/load_on_time_performance.py # Load the parquet file containing flight delay records
 on_time_dataframe = spark.read.parquet('data/on_time_performance.parquet')
 
 # Register the data for Spark SQL
 on_time_dataframe.registerTempTable("on_time_performance")
 
 # Check out the columns
 on_time_dataframe.columns
 
 # Check out some data
 on_time_dataframe
 .select("FlightDate", "TailNum", "Origin", "Dest", "Carrier", "DepDelay", "ArrDelay")
 .show()
• 34.
  Data Syndrome: AgileData Science 2.0 Sampling a DataFrame Sampling a DataFrame to get a better view of its data 34 # Trim the fields and keep the result
 trimmed_on_time = on_time_dataframe
 .select(
 "FlightDate",
 "TailNum",
 "Origin",
 "Dest",
 "Carrier",
 "DepDelay",
 "ArrDelay"
 ) # Sample 0.01% of the data and show
 trimmed_on_time.sample(False, 0.0001).show()
• 35.
  Data Syndrome: AgileData Science 2.0 Calculating a Histogram Computing the distribution of a column in a dataset 35 # See ch02/histogram.py
 # Load the parquet file containing flight delay records
 on_time_dataframe = spark.read.parquet('data/on_time_performance.parquet')
 
 # Register the data for Spark SQL
 on_time_dataframe.registerTempTable("on_time_performance")
 
 # Compute a histogram of departure delays
 on_time_dataframe
 .select("DepDelay")
 .rdd
 .flatMap(lambda x: x)
 .histogram(10)
• 36.
  Data Syndrome: AgileData Science 2.0 Displaying a Histogram Using pyplot to display a histogram 36 import numpy as np
 import matplotlib.mlab as mlab
 import matplotlib.pyplot as plt
 
 # Function to plot a histogram using pyplot
 def create_hist(rdd_histogram_data):
 """Given an RDD.histogram, plot a pyplot histogram"""
 heights = np.array(rdd_histogram_data[1])
 full_bins = rdd_histogram_data[0]
 mid_point_bins = full_bins[:-1]
 widths = [abs(i - j) for i, j in zip(full_bins[:-1], full_bins[1:])]
 bar = plt.bar(mid_point_bins, heights, width=widths, color='b')
 return bar
 
 # Compute a histogram of departure delays
 departure_delay_histogram = on_time_dataframe
 .select("DepDelay")
 .rdd
 .flatMap(lambda x: x)
 .histogram(10, [-60,-30,-15,-10,-5,0,5,10,15,30,60,90,120,180])
 
 create_hist(departure_delay_histogram)
• 37.
  Data Syndrome: AgileData Science 2.0 Displaying a Histogram Using pyplot to display a histogram 37
• 38.
  Data Syndrome: AgileData Science 2.0 Counting Airplanes How many airplanes are in the US fleet in total? 38 # See ch05/assess_airplanes.py # Load the parquet file
 on_time_dataframe = spark.read.parquet('data/on_time_performance.parquet')
 on_time_dataframe.registerTempTable("on_time_performance")
 
 # Dump the unneeded fields
 tail_numbers = on_time_dataframe.rdd.map(lambda x: x.TailNum)
 tail_numbers = tail_numbers.filter(lambda x: x != '')
 
 # distinct() gets us unique tail numbers
 unique_tail_numbers = tail_numbers.distinct()
 
 # now we need a count() of unique tail numbers
 airplane_count = unique_tail_numbers.count()
 print("Total airplanes: {}".format(airplane_count))
• 39.
  Data Syndrome: AgileData Science 2.0 Counting Total Flights by Month Preparing data for a chart 39 # See ch05/total_flights.py # Load the parquet file
 on_time_dataframe = spark.read.parquet('data/on_time_performance.parquet')
 
 # Use SQL to look at the total flights by month across 2015
 on_time_dataframe.registerTempTable("on_time_dataframe")
 total_flights_by_month = spark.sql(
 """SELECT Month, Year, COUNT(*) AS total_flights
 FROM on_time_dataframe
 GROUP BY Year, Month
 ORDER BY Year, Month"""
 )
 
 # This map/asDict trick makes the rows print a little prettier. It is optional.
 flights_chart_data = total_flights_by_month.rdd.map(lambda row: row.asDict())
 flights_chart_data.collect()
• 40.
  Data Syndrome: AgileData Science 2.0 Preparing Complex Records for Storage Getting data ready for storage in a document or key/value store 40 # See ch05/extract_airplanes.py # Load the parquet file
 on_time_dataframe = spark.read.parquet('data/on_time_performance.parquet')
 on_time_dataframe.registerTempTable("on_time_performance")
 
 # Filter down to the fields we need to identify and link to a flight
 flights = on_time_dataframe.rdd.map(lambda x: 
 (x.Carrier, x.FlightDate, x.FlightNum, x.Origin, x.Dest, x.TailNum)
 )
 
 # Group flights by tail number, sorted by date, then flight number, then origin/dest
 flights_per_airplane = flights
 .map(lambda nameTuple: (nameTuple[5], [nameTuple[0:5]]))
 .reduceByKey(lambda a, b: a + b)
 .map(lambda tuple:
 {
 'TailNum': tuple[0], 
 'Flights': sorted(tuple[1], key=lambda x: (x[1], x[2], x[3], x[4]))
 }
 )
 flights_per_airplane.first()
• 41.
  Data Syndrome: AgileData Science 2.0 Counting Flight Delays Analyzing and understanding why flights are late 41 # See ch07/explore_delays.py # Load the on-time parquet file
 on_time_dataframe = spark.read.parquet('data/on_time_performance.parquet')
 on_time_dataframe.registerTempTable("on_time_performance")
 
 total_flights = on_time_dataframe.count()
 
 # Flights that were late leaving...
 late_departures = on_time_dataframe.filter(on_time_dataframe.DepDelayMinutes > 0)
 total_late_departures = late_departures.count()
 
 # Flights that were late arriving...
 late_arrivals = on_time_dataframe.filter(on_time_dataframe.ArrDelayMinutes > 0)
 total_late_arrivals = late_arrivals.count() # Get the percentage of flights that are late, rounded to 1 decimal place
 pct_late = round((total_late_arrivals / (total_flights * 1.0)) * 100, 1)
• 42.
  Data Syndrome: AgileData Science 2.0 Hero Flights How many flights made up for time in the air? Those that departed late and arrived on time? 42 # See ch07/explore_delays.py # Flights that left late but made up time to arrive on time...
 on_time_heros = on_time_dataframe.filter(
 (on_time_dataframe.DepDelayMinutes > 0)
 &
 (on_time_dataframe.ArrDelayMinutes <= 0)
 )
 total_on_time_heros = on_time_heros.count()
• 43.
  Data Syndrome: AgileData Science 2.0 Presenting Results Displaying the answers in plaintext we’ve just calculated 43 # See ch07/explore_delays.py print("Total flights: {:,}".format(total_flights))
 print("Late departures: {:,}".format(total_late_departures))
 print("Late arrivals: {:,}".format(total_late_arrivals))
 print("Recoveries: {:,}".format(total_on_time_heros))
 print("Percentage Late: {}%".format(pct_late))
• 44.
  Data Syndrome: AgileData Science 2.0 44 # See ch07/explore_delays.py # Get the average minutes late departing and arriving
 spark.sql("""
 SELECT
 ROUND(AVG(DepDelay),1) AS AvgDepDelay,
 ROUND(AVG(ArrDelay),1) AS AvgArrDelay
 FROM on_time_performance
 """
 ).show() Average Lateness Departing and Arriving Drilling down into flights and how late they are…
• 45.
  Data Syndrome: AgileData Science 2.0 Sampling Late Flights Getting to know our data by sampling records of interest 45 # Why are flights late? Lets look at some delayed flights and the delay causes
 late_flights = spark.sql("""
 SELECT
 ArrDelayMinutes,
 WeatherDelay,
 CarrierDelay,
 NASDelay,
 SecurityDelay,
 LateAircraftDelay
 FROM
 on_time_performance
 WHERE
 WeatherDelay IS NOT NULL
 OR
 CarrierDelay IS NOT NULL
 OR
 NASDelay IS NOT NULL
 OR
 SecurityDelay IS NOT NULL
 OR
 LateAircraftDelay IS NOT NULL
 ORDER BY
 FlightDate
 """)
 late_flights.sample(False, 0.01).show()
• 46.
  Data Syndrome: AgileData Science 2.0 Why are Flights Late? Analyzing and understanding why flights are late 46 # Calculate the percentage contribution to delay for each source
 total_delays = spark.sql("""
 SELECT
 ROUND(SUM(WeatherDelay)/SUM(ArrDelayMinutes) * 100, 1) AS pct_weather_delay,
 ROUND(SUM(CarrierDelay)/SUM(ArrDelayMinutes) * 100, 1) AS pct_carrier_delay,
 ROUND(SUM(NASDelay)/SUM(ArrDelayMinutes) * 100, 1) AS pct_nas_delay,
 ROUND(SUM(SecurityDelay)/SUM(ArrDelayMinutes) * 100, 1) AS pct_security_delay,
 ROUND(SUM(LateAircraftDelay)/SUM(ArrDelayMinutes) * 100, 1) AS pct_late_aircraft_delay
 FROM on_time_performance
 """)
 total_delays.show()
• 47.
  Data Syndrome: AgileData Science 2.0 How Often are Weather Delayed Flights Late? Analyzing and understanding why flights are late 47 # Eyeball the first to define our buckets
 weather_delay_histogram = on_time_dataframe
 .select("WeatherDelay")
 .rdd
 .flatMap(lambda x: x)
 .histogram([1, 5, 10, 15, 30, 60, 120, 240, 480, 720, 24*60.0])
 print(weather_delay_histogram) create_hist(weather_delay_histogram)
• 48.
  Data Syndrome: AgileData Science 2.0 How Often are Weather Delayed Flights Late? Analyzing and understanding why flights are late 48
• 49.
  Data Syndrome: AgileData Science 2.0 Preparing Histogram Data for d3.js Analyzing and understanding why flights are late 49 # Transform the data into something easily consumed by d3
 def histogram_to_publishable(histogram):
 record = {'key': 1, 'data': []}
 for label, value in zip(histogram[0], histogram[1]):
 record['data'].append(
 {
 'label': label,
 'value': value
 }
 )
 return record # Recompute the weather histogram with a filter for on-time flights
 weather_delay_histogram = on_time_dataframe
 .filter(
 (on_time_dataframe.WeatherDelay != None)
 &
 (on_time_dataframe.WeatherDelay > 0)
 )
 .select("WeatherDelay")
 .rdd
 .flatMap(lambda x: x)
 .histogram([0, 15, 30, 60, 120, 240, 480, 720, 24*60.0])
 print(weather_delay_histogram)
 
 record = histogram_to_publishable(weather_delay_histogram)
• 50.
  Agile Data Science2.0 50 Building a classifier model Predictive Analytics Machine Learning
• 51.
  Data Syndrome: AgileData Science 2.0 Download Prepared Training Data Saving time by using a prepared dataset 51 # Be in the root directory of the project cd Agile_Data_Code_2 # Run the download script ch08/download_data.sh
• 52.
  Data Syndrome: AgileData Science 2.0 String Vectorization From properties of items to vector format 52
• 53.
  Data Syndrome: AgileData Science 2.0 53 scikit-learn was 166. Spark MLlib is very powerful! ch08/train_spark_mllib_model.py 190 Line Model # !/usr/bin/env python
 
 import sys, os, re
 
 # Pass date and base path to main() from airflow
 def main(base_path):
 
 # Default to "."
 try: base_path
 except NameError: base_path = "."
 if not base_path:
 base_path = "."
 
 APP_NAME = "train_spark_mllib_model.py"
 
 # If there is no SparkSession, create the environment
 try:
 sc and spark
 except NameError as e:
 import findspark
 findspark.init()
 import pyspark
 import pyspark.sql
 
 sc = pyspark.SparkContext()
 spark = pyspark.sql.SparkSession(sc).builder.appName(APP_NAME).getOrCreate()
 
 #
 # {
 # "ArrDelay":5.0,"CRSArrTime":"2015-12-31T03:20:00.000-08:00","CRSDepTime":"2015-12-31T03:05:00.000-08:00",
 # "Carrier":"WN","DayOfMonth":31,"DayOfWeek":4,"DayOfYear":365,"DepDelay":14.0,"Dest":"SAN","Distance":368.0,
 # "FlightDate":"2015-12-30T16:00:00.000-08:00","FlightNum":"6109","Origin":"TUS"
 # }
 #
 from pyspark.sql.types import StringType, IntegerType, FloatType, DoubleType, DateType, TimestampType
 from pyspark.sql.types import StructType, StructField
 from pyspark.sql.functions import udf
 
 schema = StructType([
 StructField("ArrDelay", DoubleType(), True), # "ArrDelay":5.0
 StructField("CRSArrTime", TimestampType(), True), # "CRSArrTime":"2015-12-31T03:20:00.000-08:00"
 StructField("CRSDepTime", TimestampType(), True), # "CRSDepTime":"2015-12-31T03:05:00.000-08:00"
 StructField("Carrier", StringType(), True), # "Carrier":"WN"
 StructField("DayOfMonth", IntegerType(), True), # "DayOfMonth":31
 StructField("DayOfWeek", IntegerType(), True), # "DayOfWeek":4
 StructField("DayOfYear", IntegerType(), True), # "DayOfYear":365
 StructField("DepDelay", DoubleType(), True), # "DepDelay":14.0
 StructField("Dest", StringType(), True), # "Dest":"SAN"
 StructField("Distance", DoubleType(), True), # "Distance":368.0
 StructField("FlightDate", DateType(), True), # "FlightDate":"2015-12-30T16:00:00.000-08:00"
 StructField("FlightNum", StringType(), True), # "FlightNum":"6109"
 StructField("Origin", StringType(), True), # "Origin":"TUS"
 ])
 
 input_path = "{}/data/simple_flight_delay_features.jsonl.bz2".format(
 base_path
 )
 features = spark.read.json(input_path, schema=schema)
 features.first()
 
 #
 # Check for nulls in features before using Spark ML
 #
 null_counts = [(column, features.where(features[column].isNull()).count()) for column in features.columns]
 cols_with_nulls = filter(lambda x: x[1] > 0, null_counts)
 print(list(cols_with_nulls))
 
 #
 # Add a Route variable to replace FlightNum
 #
 from pyspark.sql.functions import lit, concat
 features_with_route = features.withColumn(
 'Route',
 concat(
 features.Origin,
 lit('-'),
 features.Dest
 )
 )
 features_with_route.show(6)
 
 #
 # Use pysmark.ml.feature.Bucketizer to bucketize ArrDelay into on-time, slightly late, very late (0, 1, 2)
 #
 from pyspark.ml.feature import Bucketizer
 
 # Setup the Bucketizer
 splits = [-float("inf"), -15.0, 0, 30.0, float("inf")]
 arrival_bucketizer = Bucketizer(
 splits=splits,
 inputCol="ArrDelay",
 outputCol="ArrDelayBucket"
 )
 
 # Save the bucketizer
 arrival_bucketizer_path = "{}/models/arrival_bucketizer_2.0.bin".format(base_path)
 arrival_bucketizer.write().overwrite().save(arrival_bucketizer_path)
 
 # Apply the bucketizer
 ml_bucketized_features = arrival_bucketizer.transform(features_with_route)
 ml_bucketized_features.select("ArrDelay", "ArrDelayBucket").show()
 
 #
 # Extract features tools in with pyspark.ml.feature
 #
 from pyspark.ml.feature import StringIndexer, VectorAssembler
 
 # Turn category fields into indexes
 for column in ["Carrier", "Origin", "Dest", "Route"]:
 string_indexer = StringIndexer(
 inputCol=column,
 outputCol=column + "_index"
 )
 
 string_indexer_model = string_indexer.fit(ml_bucketized_features)
 ml_bucketized_features = string_indexer_model.transform(ml_bucketized_features)
 
 # Drop the original column
 ml_bucketized_features = ml_bucketized_features.drop(column)
 
 # Save the pipeline model
 string_indexer_output_path = "{}/models/string_indexer_model_{}.bin".format(
 base_path,
 column
 )
 string_indexer_model.write().overwrite().save(string_indexer_output_path)
 
 # Combine continuous, numeric fields with indexes of nominal ones
 # ...into one feature vector
 numeric_columns = [
 "DepDelay", "Distance",
 "DayOfMonth", "DayOfWeek",
 "DayOfYear"]
 index_columns = ["Carrier_index", "Origin_index",
 "Dest_index", "Route_index"]
 vector_assembler = VectorAssembler(
 inputCols=numeric_columns + index_columns,
 outputCol="Features_vec"
 )
 final_vectorized_features = vector_assembler.transform(ml_bucketized_features)
 
 # Save the numeric vector assembler
 vector_assembler_path = "{}/models/numeric_vector_assembler.bin".format(base_path)
 vector_assembler.write().overwrite().save(vector_assembler_path)
 
 # Drop the index columns
 for column in index_columns:
 final_vectorized_features = final_vectorized_features.drop(column)
 
 # Inspect the finalized features
 final_vectorized_features.show()
 
 # Instantiate and fit random forest classifier on all the data
 from pyspark.ml.classification import RandomForestClassifier
 rfc = RandomForestClassifier(
 featuresCol="Features_vec",
 labelCol="ArrDelayBucket",
 predictionCol="Prediction",
 maxBins=4657,
 maxMemoryInMB=1024
 )
 model = rfc.fit(final_vectorized_features)
 
 # Save the new model over the old one
 model_output_path = "{}/models/spark_random_forest_classifier.flight_delays.5.0.bin".format(
 base_path
 )
 model.write().overwrite().save(model_output_path)
 
 # Evaluate model using test data
 predictions = model.transform(final_vectorized_features)
 
 from pyspark.ml.evaluation import MulticlassClassificationEvaluator
 evaluator = MulticlassClassificationEvaluator(
 predictionCol="Prediction",
 labelCol="ArrDelayBucket",
 metricName="accuracy"
 )
 accuracy = evaluator.evaluate(predictions)
 print("Accuracy = {}".format(accuracy))
 
 # Check the distribution of predictions
 predictions.groupBy("Prediction").count().show()
 
 # Check a sample
 predictions.sample(False, 0.001, 18).orderBy("CRSDepTime").show(6)
 
 if __name__ == "__main__":
 main(sys.argv[1])

• 54.
  Data Syndrome: AgileData Science 2.0 Loading Our Training Data Loading our data as a DataFrame to use the Spark ML APIs 54 from pyspark.sql.types import StringType, IntegerType, FloatType, DoubleType, DateType, TimestampType
 from pyspark.sql.types import StructType, StructField
 from pyspark.sql.functions import udf
 
 schema = StructType([
 StructField("ArrDelay", DoubleType(), True), # "ArrDelay":5.0
 StructField("CRSArrTime", TimestampType(), True), # "CRSArrTime":"2015-12-31T03:20:00.000-08:00"
 StructField("CRSDepTime", TimestampType(), True), # "CRSDepTime":"2015-12-31T03:05:00.000-08:00"
 StructField("Carrier", StringType(), True), # "Carrier":"WN"
 StructField("DayOfMonth", IntegerType(), True), # "DayOfMonth":31
 StructField("DayOfWeek", IntegerType(), True), # "DayOfWeek":4
 StructField("DayOfYear", IntegerType(), True), # "DayOfYear":365
 StructField("DepDelay", DoubleType(), True), # "DepDelay":14.0
 StructField("Dest", StringType(), True), # "Dest":"SAN"
 StructField("Distance", DoubleType(), True), # "Distance":368.0
 StructField("FlightDate", DateType(), True), # "FlightDate":"2015-12-30T16:00:00.000-08:00"
 StructField("FlightNum", StringType(), True), # "FlightNum":"6109"
 StructField("Origin", StringType(), True), # "Origin":"TUS"
 ])
 
 features = spark.read.json(
 "data/simple_flight_delay_features.jsonl.bz2",
 schema=schema
 )
 features.first()
• 55.
  Data Syndrome: AgileData Science 2.0 Checking the Data for Nulls Nulls will cause problems hereafter, so detect and address them first 55 #
 # Check for nulls in features before using Spark ML
 #
 null_counts = [(column, features.where(features[column].isNull()).count()) for column in features.columns]
 cols_with_nulls = filter(lambda x: x[1] > 0, null_counts)
 print(list(cols_with_nulls))
• 56.
  Data Syndrome: AgileData Science 2.0 Adding a Feature - The Route Route is defined as origin airport code + “-“ + destination airport code 56 #
 # Add a Route variable to replace FlightNum
 #
 from pyspark.sql.functions import lit, concat
 
 features_with_route = features.withColumn(
 'Route',
 concat(
 features.Origin,
 lit('-'),
 features.Dest
 )
 )
 features_with_route.select("Origin", "Dest", "Route").show(5)
• 57.
  Data Syndrome: AgileData Science 2.0 Bucketizing ArrDelay into ArrDelayBucket We can’t classify a continuous variable, so we must bucketize it to make it nominal/categorical 57 #
 # Use pysmark.ml.feature.Bucketizer to bucketize ArrDelay
 #
 from pyspark.ml.feature import Bucketizer
 
 splits = [-float("inf"), -15.0, 0, 30.0, float("inf")]
 bucketizer = Bucketizer(
 splits=splits,
 inputCol="ArrDelay",
 outputCol="ArrDelayBucket"
 )
 ml_bucketized_features = bucketizer.transform(features_with_route)
 
 # Check the buckets out
 ml_bucketized_features.select("ArrDelay", "ArrDelayBucket").show()
• 58.
  Data Syndrome: AgileData Science 2.0 Indexing String Columns into Numeric Columns Nominal/categorical/string columns need to be made numeric before we can vectorize them 58 #
 # Extract features tools in with pyspark.ml.feature
 #
 from pyspark.ml.feature import StringIndexer, VectorAssembler
 
 # Turn category fields into categoric feature vectors, then drop intermediate fields
 for column in ["Carrier", "DayOfMonth", "DayOfWeek", "DayOfYear",
 "Origin", "Dest", "Route"]:
 string_indexer = StringIndexer(
 inputCol=column,
 outputCol=column + "_index"
 )
 ml_bucketized_features = string_indexer.fit(ml_bucketized_features)
 .transform(ml_bucketized_features)
 
 # Check out the indexes
 ml_bucketized_features.show(6)
• 59.
  Data Syndrome: AgileData Science 2.0 Combining Numeric and Indexed Fields into One Vector Our classifier needs a single field, so we combine all our numeric fields into one feature vector 59 # Handle continuous, numeric fields by combining them into one feature vector
 numeric_columns = ["DepDelay", "Distance"]
 index_columns = ["Carrier_index", "DayOfMonth_index",
 "DayOfWeek_index", "DayOfYear_index", "Origin_index",
 "Origin_index", "Dest_index", "Route_index"]
 vector_assembler = VectorAssembler(
 inputCols=numeric_columns + index_columns,
 outputCol="Features_vec"
 )
 final_vectorized_features = vector_assembler.transform(ml_bucketized_features)
 
 # Drop the index columns
 for column in index_columns:
 final_vectorized_features = final_vectorized_features.drop(column)
 
 # Check out the features
 final_vectorized_features.show()
• 60.
  Data Syndrome: AgileData Science 2.0 Splitting our Data in a Test/Train Split We need to split our data to evaluate the performance of our classifier 60 #
 # Cross validate, train and evaluate classifier
 #
 
 # Test/train split
 training_data, test_data = final_vectorized_features.randomSplit([0.7, 0.3])
• 61.
  Data Syndrome: AgileData Science 2.0 Training Our Model This is the magic in machine learning, and it is only a couple of lines of code 61 # Instantiate and fit random forest classifier
 from pyspark.ml.classification import RandomForestClassifier
 rfc = RandomForestClassifier(
 featuresCol="Features_vec",
 labelCol="ArrDelayBucket",
 maxBins=4657,
 maxMemoryInMB=1024
 )
 model = rfc.fit(training_data)
• 62.
  Data Syndrome: AgileData Science 2.0 Evaluating Our Model Using the test/train split to evaluate our model for accuracy 62 # Evaluate model using test data
 predictions = model.transform(test_data)
 
 from pyspark.ml.evaluation import MulticlassClassificationEvaluator
 evaluator = MulticlassClassificationEvaluator(labelCol="ArrDelayBucket", metricName="accuracy")
 accuracy = evaluator.evaluate(predictions)
 print("Accuracy = {}".format(accuracy))
• 63.
  Data Syndrome: AgileData Science 2.0 Sampling Our Predictions Making sure they pass the sniff check 63 # Check a sample
 predictions.sample(False, 0.001, 18).orderBy("CRSDepTime").show(6)
• 64.
  Data Syndrome: AgileData Science 2.0 Experiment Setup Necessary to improve model 64
• 65.
  Data Syndrome: AgileData Science 2.0 65 155 additional lines to setup an experiment and add 3 new features to improvement the model ch09/improve_spark_mllib_model.py 345 L.O.C. # !/usr/bin/env python
 
 import sys, os, re
 import json
 import datetime, iso8601
 from tabulate import tabulate
 
 # Pass date and base path to main() from airflow
 def main(base_path):
 APP_NAME = "train_spark_mllib_model.py"
 
 # If there is no SparkSession, create the environment
 try:
 sc and spark
 except NameError as e:
 import findspark
 findspark.init()
 import pyspark
 import pyspark.sql
 
 sc = pyspark.SparkContext()
 spark = pyspark.sql.SparkSession(sc).builder.appName(APP_NAME).getOrCreate()
 
 #
 # {
 # "ArrDelay":5.0,"CRSArrTime":"2015-12-31T03:20:00.000-08:00","CRSDepTime":"2015-12-31T03:05:00.000-08:00",
 # "Carrier":"WN","DayOfMonth":31,"DayOfWeek":4,"DayOfYear":365,"DepDelay":14.0,"Dest":"SAN","Distance":368.0,
 # "FlightDate":"2015-12-30T16:00:00.000-08:00","FlightNum":"6109","Origin":"TUS"
 # }
 #
 from pyspark.sql.types import StringType, IntegerType, FloatType, DoubleType, DateType, TimestampType
 from pyspark.sql.types import StructType, StructField
 from pyspark.sql.functions import udf
 
 schema = StructType([
 StructField("ArrDelay", DoubleType(), True), # "ArrDelay":5.0
 StructField("CRSArrTime", TimestampType(), True), # "CRSArrTime":"2015-12-31T03:20:00.000-08:00"
 StructField("CRSDepTime", TimestampType(), True), # "CRSDepTime":"2015-12-31T03:05:00.000-08:00"
 StructField("Carrier", StringType(), True), # "Carrier":"WN"
 StructField("DayOfMonth", IntegerType(), True), # "DayOfMonth":31
 StructField("DayOfWeek", IntegerType(), True), # "DayOfWeek":4
 StructField("DayOfYear", IntegerType(), True), # "DayOfYear":365
 StructField("DepDelay", DoubleType(), True), # "DepDelay":14.0
 StructField("Dest", StringType(), True), # "Dest":"SAN"
 StructField("Distance", DoubleType(), True), # "Distance":368.0
 StructField("FlightDate", DateType(), True), # "FlightDate":"2015-12-30T16:00:00.000-08:00"
 StructField("FlightNum", StringType(), True), # "FlightNum":"6109"
 StructField("Origin", StringType(), True), # "Origin":"TUS"
 ])
 
 input_path = "{}/data/simple_flight_delay_features.json".format(
 base_path
 )
 features = spark.read.json(input_path, schema=schema)
 features.first()
 
 #
 # Add a Route variable to replace FlightNum
 #
 from pyspark.sql.functions import lit, concat
 features_with_route = features.withColumn(
 'Route',
 concat(
 features.Origin,
 lit('-'),
 features.Dest
 )
 )
 features_with_route.show(6)
 
 #
 # Add the hour of day of scheduled arrival/departure
 #
 from pyspark.sql.functions import hour
 features_with_hour = features_with_route.withColumn(
 "CRSDepHourOfDay",
 hour(features.CRSDepTime)
 )
 features_with_hour = features_with_hour.withColumn(
 "CRSArrHourOfDay",
 hour(features.CRSArrTime)
 )
 features_with_hour.select("CRSDepTime", "CRSDepHourOfDay", "CRSArrTime", "CRSArrHourOfDay").show()
 
 #
 # Use pysmark.ml.feature.Bucketizer to bucketize ArrDelay into on-time, slightly late, very late (0, 1, 2)
 #
 from pyspark.ml.feature import Bucketizer
 
 # Setup the Bucketizer
 splits = [-float("inf"), -15.0, 0, 30.0, float("inf")]
 arrival_bucketizer = Bucketizer(
 splits=splits,
 inputCol="ArrDelay",
 outputCol="ArrDelayBucket"
 )
 
 # Save the model
 arrival_bucketizer_path = "{}/models/arrival_bucketizer_2.0.bin".format(base_path)
 arrival_bucketizer.write().overwrite().save(arrival_bucketizer_path)
 
 # Apply the model
 ml_bucketized_features = arrival_bucketizer.transform(features_with_hour)
 ml_bucketized_features.select("ArrDelay", "ArrDelayBucket").show()
 
 #
 # Extract features tools in with pyspark.ml.feature
 #
 from pyspark.ml.feature import StringIndexer, VectorAssembler
 
 # Turn category fields into indexes
 for column in ["Carrier", "Origin", "Dest", "Route"]:
 string_indexer = StringIndexer(
 inputCol=column,
 outputCol=column + "_index"
 )
 
 string_indexer_model = string_indexer.fit(ml_bucketized_features)
 ml_bucketized_features = string_indexer_model.transform(ml_bucketized_features)
 # Save the pipeline model
 string_indexer_output_path = "{}/models/string_indexer_model_3.0.{}.bin".format(
 base_path,
 column
 )
 string_indexer_model.write().overwrite().save(string_indexer_output_path)
 
 # Combine continuous, numeric fields with indexes of nominal ones
 # ...into one feature vector
 numeric_columns = [
 "DepDelay", "Distance",
 "DayOfMonth", "DayOfWeek",
 "DayOfYear", "CRSDepHourOfDay",
 "CRSArrHourOfDay"]
 index_columns = ["Carrier_index", "Origin_index",
 "Dest_index", "Route_index"]
 vector_assembler = VectorAssembler(
 inputCols=numeric_columns + index_columns,
 outputCol="Features_vec"
 )
 final_vectorized_features = vector_assembler.transform(ml_bucketized_features)
 
 # Save the numeric vector assembler
 vector_assembler_path = "{}/models/numeric_vector_assembler_3.0.bin".format(base_path)
 vector_assembler.write().overwrite().save(vector_assembler_path)
 
 # Drop the index columns
 for column in index_columns:
 final_vectorized_features = final_vectorized_features.drop(column)
 
 # Inspect the finalized features
 final_vectorized_features.show()
 
 #
 # Cross validate, train and evaluate classifier: loop 5 times for 4 metrics
 #
 
 from collections import defaultdict
 scores = defaultdict(list)
 feature_importances = defaultdict(list)
 metric_names = ["accuracy", "weightedPrecision", "weightedRecall", "f1"]
 split_count = 3
 
 for i in range(1, split_count + 1):
 print("nRun {} out of {} of test/train splits in cross validation...".format(
 i,
 split_count,
 )
 )
 
 # Test/train split
 training_data, test_data = final_vectorized_features.randomSplit([0.8, 0.2])
 
 # Instantiate and fit random forest classifier on all the data
 from pyspark.ml.classification import RandomForestClassifier
 rfc = RandomForestClassifier(
 featuresCol="Features_vec",
 labelCol="ArrDelayBucket",
 predictionCol="Prediction",
 maxBins=4657,
 )
 model = rfc.fit(training_data)
 
 # Save the new model over the old one
 model_output_path = "{}/models/spark_random_forest_classifier.flight_delays.baseline.bin".format(
 base_path
 )
 model.write().overwrite().save(model_output_path)
 
 # Evaluate model using test data
 predictions = model.transform(test_data)
 
 # Evaluate this split's results for each metric
 from pyspark.ml.evaluation import MulticlassClassificationEvaluator
 for metric_name in metric_names:
 evaluator = MulticlassClassificationEvaluator(
 labelCol="ArrDelayBucket",
 predictionCol="Prediction",
 metricName=metric_name
 )
 score = evaluator.evaluate(predictions)
 
 scores[metric_name].append(score)
 print("{} = {}".format(metric_name, score))
 
 #
 # Collect feature importances
 #
 feature_names = vector_assembler.getInputCols()
 feature_importance_list = model.featureImportances
 for feature_name, feature_importance in zip(feature_names, feature_importance_list):
 feature_importances[feature_name].append(feature_importance)
 
 #
 # Evaluate average and STD of each metric and print a table
 #
 import numpy as np
 score_averages = defaultdict(float)
 
 # Compute the table data
 average_stds = [] # ha
 for metric_name in metric_names:
 metric_scores = scores[metric_name]
 
 average_accuracy = sum(metric_scores) / len(metric_scores)
 score_averages[metric_name] = average_accuracy
 
 std_accuracy = np.std(metric_scores)
 
 average_stds.append((metric_name, average_accuracy, std_accuracy))
 
 # Print the table
 print("nExperiment Log")
 print("--------------")
 print(tabulate(average_stds, headers=["Metric", "Average", "STD"]))
 
 #
 # Persist the score to a sccore log that exists between runs
 #
 import pickle # Load the score log or initialize an empty one
 try:
 score_log_filename = "{}/models/score_log.pickle".format(base_path)
 score_log = pickle.load(open(score_log_filename, "rb"))
 if not isinstance(score_log, list):
 score_log = []
 except IOError:
 score_log = []
 
 # Compute the existing score log entry
 score_log_entry = {metric_name: score_averages[metric_name] for metric_name in metric_names}
 
 # Compute and display the change in score for each metric
 try:
 last_log = score_log[-1]
 except (IndexError, TypeError, AttributeError):
 last_log = score_log_entry
 
 experiment_report = []
 for metric_name in metric_names:
 run_delta = score_log_entry[metric_name] - last_log[metric_name]
 experiment_report.append((metric_name, run_delta))
 
 print("nExperiment Report")
 print("-----------------")
 print(tabulate(experiment_report, headers=["Metric", "Score"]))
 
 # Append the existing average scores to the log
 score_log.append(score_log_entry)
 
 # Persist the log for next run
 pickle.dump(score_log, open(score_log_filename, "wb"))
 
 #
 # Analyze and report feature importance changes
 #
 
 # Compute averages for each feature
 feature_importance_entry = defaultdict(float)
 for feature_name, value_list in feature_importances.items():
 average_importance = sum(value_list) / len(value_list)
 feature_importance_entry[feature_name] = average_importance
 
 # Sort the feature importances in descending order and print
 import operator
 sorted_feature_importances = sorted(
 feature_importance_entry.items(),
 key=operator.itemgetter(1),
 reverse=True
 )
 
 print("nFeature Importances")
 print("-------------------")
 print(tabulate(sorted_feature_importances, headers=['Name', 'Importance']))
 
 #
 # Compare this run's feature importances with the previous run's
 #
 
 # Load the feature importance log or initialize an empty one
 try:
 feature_log_filename = "{}/models/feature_log.pickle".format(base_path)
 feature_log = pickle.load(open(feature_log_filename, "rb"))
 if not isinstance(feature_log, list):
 feature_log = []
 except IOError:
 feature_log = []
 
 # Compute and display the change in score for each feature
 try:
 last_feature_log = feature_log[-1]
 except (IndexError, TypeError, AttributeError):
 last_feature_log = defaultdict(float)
 for feature_name, importance in feature_importance_entry.items():
 last_feature_log[feature_name] = importance
 
 # Compute the deltas
 feature_deltas = {}
 for feature_name in feature_importances.keys():
 run_delta = feature_importance_entry[feature_name] - last_feature_log[feature_name]
 feature_deltas[feature_name] = run_delta
 
 # Sort feature deltas, biggest change first
 import operator
 sorted_feature_deltas = sorted(
 feature_deltas.items(),
 key=operator.itemgetter(1),
 reverse=True
 )
 
 # Display sorted feature deltas
 print("nFeature Importance Delta Report")
 print("-------------------------------")
 print(tabulate(sorted_feature_deltas, headers=["Feature", "Delta"]))
 
 # Append the existing average deltas to the log
 feature_log.append(feature_importance_entry)
 
 # Persist the log for next run
 pickle.dump(feature_log, open(feature_log_filename, "wb"))
 
 if __name__ == "__main__":
 main(sys.argv[1])

• 66.
  Data Syndrome RussellJurney Principal Consultant Email : [email protected] Web : datasyndrome.com Data Syndrome, LLC