CSE6242 - Data and Visual Analytics - HW 3 -Spark -Docker - DataBricks - AWS and GCP  - Solved

Homework Overview
Many modern-day datasets are huge and truly exemplify “big data”. For example, the Facebook social graph is petabytes large (over 1M GB); every day, Twitter users generate over 12 terabytes of messages; and the NASA Terra and Aqua satellites each produce over 300 GB of MODIS satellite imagery per day. These raw data are far too large to even fit on the hard drive of an average computer, let alone to process and analyze. Luckily, there are a variety of modern technologies that allow us to process and analyze such large datasets in a reasonable amount of time. For the bulk of this assignment, you will be working with a dataset of over 1 billion individual taxi trips from the New York City Taxi & Limousine Commission (TLC). Further details on this dataset are available here.

In Q1, you will work with a subset of the TLC dataset to get warmed up with PySpark. Apache Spark is a framework for distributed computing, and PySpark is its Python API. You will use this tool to answer questions such as “what are the top 10 most common trips in the dataset”? You will be using your own machine for computation, using an environment defined by a Docker container.

In Q2, you will perform further analysis on a different subset of the TLC dataset using Spark on DataBricks, a platform combining datasets, machine learning models, and cloud compute. This part of the assignment will be completed in the Scala programming language, a modern general-purpose language with a robust support for functional programming. The Spark distributed computing framework is in fact written using Scala.

In Q3, you will use PySpark on AWS using Elastic MapReduce (EMR), and in Q4 you will use Spark on Google Cloud Platform, to analyze even larger samples from the TLC dataset.

Finally, in Q5 you will use the Microsoft Azure ML Studio to implement a regression model to predict automobile prices using a sample dataset already included in the Azure workspace. A main goal of this assignment is to help students gain exposure to a variety of tools that will be useful in the future (e.g., future project, research, career). The reasoning behind intentionally including AWS, Azure and GCP (most courses use only one), because we want students to be able to try and compare these platforms as they evolve rapidly. This will help the students in the future should they need to select a cloud platform to use, they can make more informed decisions and be able to get started right away.  

You will find that a number of computational tasks in this assignment are not very difficult, and there seems to be quite a bit of “setup” to do before getting to the actual “programming” part of the problem. The reasoning behind this design is because for many students, this assignment is the very first time they use any cloud services; they are new to the pay-per-use model, and they have never used a cluster of machines. There are over 1000 students in CSE 6242 (campus and online) and CX 4242 combined. This means we have students coming from a great variety of backgrounds. We wished we could provide every student unlimited AWS credit so that they can try out many services and write programs that are more complex. Over the past offering of this course, we have been gradually increasing the “programming” part and reducing much of the “setup” (e.g., the use of Docker, Databricks and Jupyter notebooks were major improvements). We will continue to further reduce the setup that students need to perform in future offerings of this course.

Q1 [15 points] Analyzing trips data with PySpark
Follow these instructions to download and setup a preconfigured Docker image that you will use for this assignment.  

 

Why use Docker? In earlier iterations of this course, students installed software on their own machines, and we (both students and instructor team) ran into all sorts of issues, some of which could not be resolved satisfactorily. Docker allows us to distribute a cross platform, preconfigured image with all of the requisite software and correct package versions. Once Docker is installed and the container is running, access Jupyter by browsing to http://localhost:6242. There is no need to install any additional Java or PySpark dependencies as they are all bundled as part of the Docker container.  

 

Imagine that your boss gives you a large dataset which contains trip information of New York City Taxi and Limousine Commission (TLC). You are asked to provide summaries for the most common trips, as well as information related to fares and traffic. This information might help in positioning taxis depending on the demand at each location.

 

You are provided with a Jupyter notebook (q1.ipynb) file which you will complete using PySpark using the provided Docker image. Be sure to save your work often! If you do not see your notebook in Jupyter then double-check that the file is present in the folder and double check that your Docker has been set up correctly. If, after checking both, the file still does not appear in Jupyter then you can still move forward by clicking the “upload” button in the Jupyter notebook and uploading the file – however, if you use this approach then your file will not be saved to disk when you save in Jupyter, so you would need to download your work by going to File > Download as... > Notebook (.ipynb), so be sure to download as often as you would normally save!

Note:

1.    Regular PySpark Dataframe Operations and PySpark SQL operations can be used. 

2.    If you re-run cells, remember to restart the kernel to clear the Spark context, otherwise an existing Spark context may cause errors. 

 

Tasks
You will use the yellow_tripdata_2019-01_short.csv dataset. This dataset is a modified record of the NYC Green Taxi trips and includes information about the pick-up and drop-off dates/times, pick-up and drop-off locations, trip distances, fare amounts, payment types, and driver-reported passenger counts. When processing the data or performing calculations, do not round any values. Download the data here.

 

a.    [1 pt] You will be modifying the function clean_data to clean the data. Cast the following columns into the specified data types: 

a.    passenger_count - integer

b.    total_amount - float

c.     tip_amount - float

d.    trip_distance - float

e.    fare_amount - float

f.      tpep_pickup_datetime - timestamp

g.    tpep_dropoff_datetime – timestamp

 

b.    [4 pts] You will be modifying the function common_pair. Find the top 10 pickup-dropoff location pairs having the highest number of trips (count). The location pairs should be ordered by count in descending order. If two or more pairs have the same number of trips, break the tie using the trip amount per distance travelled (trip_rate) in descending order. Use columns total_amount and trip_distance to calculate the trip amount per distance. In certain situations, the pickup and dropoff locations may be the same (include such entries as well).  

 

While calculating trip_rate, first get the average trip_distance and the average total_amount for each pair of PULocationID and DOLocationID (using group by). Then take their ratio to get the trip_rate for a pickup-drop pair. 

 

Example:

 

   

Sample Output: 

PULocationID
DOLocationID
Count 
trip_rate
1 
2
23 
5.242345 
3 
3
5 
6.61345634 
 

c.    [4 pts] You will be modifying the function time_of_cheapest_fare. Divide each day into two periods: Day (from 9am to 8:59:59pm, both inclusive), and Night (from 9pm to 8:59:59am, both inclusive). Calculate the average total amount per unit distance travelled (use column total_amount) for both time periods. Sort the result by trip_rate in ascending order to determine when the fare rate is the cheapest. Use tpep_pickup_datetime to divide trips into Day and Night. Output:

day_night
trip_rate 
Day 
4.2632344561 
Night 
6.42342882 
 

d.    [4 pts] You will be modifying the function passenger_count_for_most_tip . Filter the data for trips having fares (fare_amount) greater than $2 and the number of passengers

(passenger_count) greater than 0. Calculate the average fare and tip (tip_amount) for all passenger group sizes and calculate the tip percent (tip_amount * 100 / fare_amount). Sort the result in descending order of tip percent to obtain the group size that tips the most generously.

 

Output:  

passenger_count
tip_percent
2 
14.22345234 
1 
12.523334576 
3 
12.17345231 
 

e.    [3 pts] You will be modifying the function day_with_traffic . Sort the days of the week (using tpep_pickup_datetime) in descending order of traffic (day having the highest traffic should be at the top). Calculate traffic for a particular day using the average speed of all taxi trips on that day of the week. Calculate the average speed as the average trip distance divided by the average trip time, as distance per hour. If the average_speed is equal for multiple days, order the days

alphabetically. A day with low average speed indicates high levels of traffic. The average speed may be 0, indicating very high levels of traffic. Not all days of the week may be present in the data (do not include these missing days of the week in your output). Use date_format along with the appropriate pattern letters to format the day of the week such that it matches the example output below.  

 

Output:  

day_of_week
average_speed
Fri 
0.953452345 
Mon 
5.2424622 
Tue 
9.23345272 

Q2 [30 pts] Analyzing dataset with Spark/Scala on Databricks
Tutorial 

Firstly, go over this Spark on Databricks Tutorial, to learn the basics of creating Spark jobs, loading data, and working with data.

 

You will analyze nyc-tripdata.csv[1] using Spark and Scala on the Databricks platform. (A short description of how Spark and Scala are related can be found here.) You will also need to use the taxi zone lookup table using taxi_zone_lookup.csv that maps the location ID into the actual name of the region in NYC. The nyctripdata dataset is a modified record of the NYC Green Taxi trips and includes information about the pick-up and drop-off dates/times, pick-up and drop-off locations, trip distances, fare amounts, payment types, and driver-reported passenger counts.

VERY IMPORTANT  
1.    Use only Firefox, Safari or Chrome when configuring anything related to Databricks. The setup process has been verified to work on these browsers.

2.    Carefully follow the instructions in the Databricks Setup Guide (Datasets mentioned above must be downloaded from here. This link has been mentioned in the guide as well.). Open the link in a private browser window if you get ‘Permission Denied’ message.  

a.    You must choose the Databricks Runtime (DBR) version as “6.4 (includes Apache Spark 2.4.5, Scala 2.11)”. 

b.    You must not choose the default DBR version of >= 7.2

c.     Note that you do not need to install Scala or spark in your local machine. They are provided with the DBR environment.

3.    You must use only Scala DataFrame operations for this question. Scala DataFrames are just another name for Spark DataSet of rows. You can use DataSet API in Spark to work on these DataFrames. Here is a Spark document that will help you get started on working with DataFrames in Spark. You will lose points if you use SQL queries, Python, or R to manipulate a DataFrame.  

a.    Refer to this link to understand how to avoid using other languages. After selecting the default language as SCALA, do not use the language magic %<language> with other languages like %r, %python, %sql etc. The language magics are used to override the default language, which you must not do for this assignment.  

b.    You must not use full SQL queries in lieu of the Spark DataFrame API. That is, you must not use functions like sql(), which allows you to directly write full SQL queries like spark.sql (“SELECT* FROM col1 WHERE …”).  This should be df.select(“*”) instead. 

4.    The template Scala notebook q2.dbc (in hw3-skeleton) provides you with code that reads a data file nyc-tripdata.csv. The input data is loaded into a DataFrame, inferring the schema using reflection (Refer to the Databricks Setup Guide above). It also contains code that filters the data to only keep the rows where the pickup location is different from the drop location, and the trip distance is strictly greater than 2.0 (>2.0).  

a.    All tasks listed below must be performed on this filtered DataFrame, or you will end up with wrong answers.  

 

b.    Carefully read the instructions in the notebook, which also provide hints for solving the problems.

 

5.    Some tasks in this question have specified data types for the results that are of lower precisions (e.g., float). For these tasks, we will accept relevant higher precision formats (e.g., double). Similarly, we will accept results stored in data types that offer “greater range” (e.g., long, bigint) than what we have specified (e.g., int). 

Tasks 
1)    List the top-5 most popular locations for:

a.    [2 pts] dropoff based on "DOLocationID", sorted in descending order. If there is a tie, then one with a lower "DOLocationID" gets listed first. 

b.    [2 pts] pickup based on "PULocationID", sorted in descending order. If there is a tie, then one with a lower "PULocationID" gets listed first. 

2)    [4 pts] List the top-3 locationID’s with the maximum overall activity. Here, overall activity at a LocationID is simply the sum of all pickups and all drop offs at that LocationID. In case of a tie, the lower LocationID gets listed first.

Note: If a taxi picked up 3 passengers at once, we count it as 1 pickup and not 3 pickups.

 

3)    [4 pts] List all the boroughs (of NYC: Manhattan, Brooklyn, Queens, Staten Island, Bronx along with "Unknown" and "EWR") and their total number of activities, in descending order of total number of activities. Here, the total number of activities for a borough (e.g., Queens) is the sum of the overall activities (as defined in part 2) of all the LocationIDs that fall in that borough (Queens). An example output is shown below.

  

4)    [5 pts] List the top 2 days of week with the largest number of (daily) average pickups, along with the values of the average number of pickups on each of the two days in descending order. Here, the average pickup is calculated by taking an average of the number of pickups on different dates falling on the same day of the week. For example, 02/01/2021, 02/08/2021 and 02/15/2021 are all Mondays, so the average pickups for these is the sum of the pickups on each date divided by 3. An example output is shown below.

  

Note: The day of week should be a string with its full name, for example, "Monday" instead of a              number 1 or "Mon".

5)    [6 pts] For each particular hour of a day (0 to 23, 0 being midnight) — in their order from 0 to 23 (inclusively), find the zone in the Brooklyn borough with the largest number of pickups.

6)    [7 pts] Find which 3 different days in the month of January, in Manhattan, that saw the largest percentage increment in pickups compared to previous day, in the order from largest increment % to smallest increment %. An example output is shown below.

  

 

List the results of the above tasks in the provided q2_results.csv file under the relevant sections. These preformatted sections also show you the required output format from your Scala code with the necessary columns — while column names can be different, their resulting values must be correct.  

•       You must manually enter the output generated into the corresponding sections of the q2_results.csv file, preferably using some spreadsheet software like MS-Excel (but make sure to keep the csv format). For generating the output in the Scala notebook, refer to show() and display()functions of Scala.  

•       Note that you can edit this csv file using text editor, but please be mindful about putting the results under designated columns.  

Note: Do NOT modify anything other than filling in those required output values in this csv file. We grade by running the Spark Scala code you write and by looking at your results listed in this file. So, make sure that your output is actually obtained from the Spark Scala code you write.

Hint: You may find some of the following DataFrame operations helpful:  

toDF, join, select, groupBy, orderBy, filter, agg, Window(), partitionBy, orderBy, etc.

 

Q3 [35 points] Analyzing Large Amount of Data with PySpark on AWS
VERY IMPORTANT: Use Firefox, Safari or Chrome when configuring anything related to AWS. 

 

You will try out PySpark for processing data on Amazon Web Services (AWS). Here you can learn more about PySpark and how it can be used for data analysis. You will be completing a task that may be accomplished using a commodity computer (e.g., consumer-grade laptops or desktops). However, we would like you to use this exercise as an opportunity to learn distributed computing on Amazon EC2, and to gain experience that will help you tackle more complex problems.

 

The services you will primarily be using are Amazon S3 storage, Amazon Elastic Cloud Computing (EC2) virtual servers, and Amazon Elastic MapReduce (EMR) managed Hadoop framework. You will be creating an S3 bucket, running code through EMR, and then storing the output into that S3 bucket.

 

For this question, you will only use up a very small fraction of your AWS credit.  

 

If you have any issues with the AWS credits or educate account, please fill out this form. 

AWS Guidelines
Please read the AWS Setup Tutorial to set up your AWS account. Instructions are provided both as a written guide, and a video tutorial. 

Datasets
In this question, you will use a dataset of trip records provided by the New York City Taxi and Limousine Commission (TLC). Further details on this dataset are available here and here.  From these pages [1] [2], you can explore the structure of the data, however you will be accessing the dataset directly through AWS via the code outlined in the homework skeleton. You will be working with two samples of this data, one small, and one much larger.

 

EXTREMELY IMPORTANT: Both the datasets are in the US East (N. Virginia) region. Using machines in other regions for computation would incur data transfer charges. Hence, set your region to US East (N. Virginia) in the beginning (not Oregon, which is the default). This is extremely important, otherwise your code may not work, and you may be charged extra.

Goal
You work at NYC TLC, and since the company bought a few new taxis, your boss has asked you to locate potential places where taxi drivers can pick up more passengers. Of course, the more profitable the locations are, the better. Your boss also tells you not to worry about short trips for any of your analysis, so only analyze trips which are 2.0 miles or longer.

 

First, find the 20 most popular drop off locations in the Manhattan borough by finding which of these destinations had the greatest passenger count.

 

Now, analyze all pickup locations, regardless of borough.  

•       For each pickup location determine  o the average total amount per trip,  

o   the total count of all trips that start at that location, and  

o   the count of all trips that start at that location and end at one of most popular drop off locations.  

•       Using the above values,  o determine the proportion of trips that end in one of the popular drop off locations (# trips that end in drop off location divided by total # of trips) and  

o   multiply that proportion by the average total amount to get a weighted profit value based on the probability of passengers going to one of the popular destinations.

 

Bear in mind, your boss is not as savvy with the data as you are and is not interested in location IDs. To make it easy for your boss, provide the Borough and Zone for each of the top 20 pickup locations you determined. To help you evaluate the correctness of your output, we have provided you with the output for the small dataset. Keep in mind that the small dataset and its output can be thought of as only a single “test case” for the large dataset and cannot test for all possible scenarios for the large dataset. That is, running code on the small dataset and producing expected results does NOT necessarily mean the same code will produce the correct results for the large dataset.  

 

Tasks
You are provided with a python notebook (q3_pyspark.ipynb) file which you will complete and load into EMR. You are provided with the load_data() function, which loads two PySpark DataFrames. The first is trips which contains a DataFrame of trip data, where each record refers to one (1) trip. The second is lookup which maps a LocationID to its information. It can be linked to either the PULocationID or DOLocationID fields in the trips DataFrame.

 

The following functions must be completed for full credit.  

 

VERY IMPORTANT 

•       Ensure that the parameters for each function remain as defined and the output order and names of the fields in the PySpark DataFrames are maintained.  

•       Do not import any functions which were not already imported within the skeleton. 

•       You must NOT round any numeric values. Rounding numbers can introduce inaccuracies. Our grader will be checking the first 8 decimal places of each value in the DataFrame.  

 

a)    [1 pts] user()

i.           Returns your GT Username as a string (e.g., gburdell3)

b)    [2 pts] long_trips(trips)

i.           This function filters trips to keep only trips 2 miles or longer (e.g., >= 2).

ii.         Returns PySpark DataFrame with the same schema as trips

iii.        Note: Parts c, d and e will use the result of this function 

c)    [6 pts] manhattan_trips(trips, lookup)

i.           This function determines the top 20 locations with a DOLocationID in Manhattan by sum of passenger count.

ii.         Returns a PySpark DataFrame with the schema (DOLocationID, pcount)

d)    [6 pts] weighted_profit(trips, mtrips)

i.           This function determines  

i.      the average total_amount,  

ii.    the total count of trips, and  iii. the total count of trips ending in the top 20 destinations iv.  and return the weighted_profit as discussed earlier in the homework document.

v. Returns a PySpark DataFrame with the schema (PULocationID, weighted_profit) for the weighted_profit as discussed earlier in this homework document.

e)    [5 pts] final_output(wp, lookup)

i.           This function  

i.      takes the results of weighted_profit,  

ii.    links it to the borough and zone through the lookup data frame, and  

iii.   returns the top 20 locations with the highest weighted_profit.

ii.         Returns a PySpark DataFrame with the schema (Zone, Borough, weighted_profit)

 

Once you have implemented all these functions, run the main() function, which is already implemented, and update the line of code to include the name of your output s3 bucket and a location. This function will fail if the output directory already exists, so make sure to change it each time you run the function.

 

Example: final.write.csv('s3://cse6242-bburdell3/output-large3) 

 

Your output file will appear in a folder in your s3 bucket as a csv file with a name which is similar to part0000-4d992f7a-0ad3-48f8-8c72-0022984e4b50-c000.csv. Download this file and rename it to q3_output.csv . Do not make any other changes to the file.

 

Hint: Refer to commands such as filter, join, groupBy, agg, limit, sort, withColumnRenamed and withColumn.  

 

 


Q4 [10 points] Analyzing a Large Dataset using Spark on GCP 

VERY IMPORTANT: Use Firefox, Safari or Chrome when configuring anything related to GCP. 

GCP Guidelines
Instructions to setup GCP Credits, GCP Storage and Dataproc Cluster are provided as video tutorials (here, here and here) and as written instructions 

 

Helpful tips/FAQs for special scenarios: 

 

a)  If GCP service is disabled for your google account, try the steps in this google support link  

b)  If you have any issues with GCP free credits, please fill out this form 

Goal
The goal of this question is to familiarize you with creating storage buckets/clusters and running Spark programs on Google Cloud Platform. This question asks you to create a new Google Storage Bucket and load the NYC Taxi & Limousine Commission Dataset. You are also provided with a Jupyter Notebook (q4_pyspark-gcp.ipynb) file which you will load and complete in Google Dataproc Cluster. Inside the notebook, you are provided with the load_data() function, which you will complete to load a PySpark DataFrame from the Google Storage Bucket you created as part of this question. Using this PySpark DataFrame, you will complete the following tasks using Spark DataFrame functions or Spark SQL functions. 

 

 

You will use the data file yellow_tripdata_2019-01.csv. Each line represents a single taxi trip consisting of following comma separated columns. All columns are of string data type. You must convert the highlighted columns below into decimal data type (do NOT use float datatype) inside the load_data function or individual functions when completing this question: 

•       VendorID 

•       tpep_pickup_datetime 

•       tpep_dropoff_datetime 

•       passenger_count 

•       trip_distance (decimal data type) 

•       RatecodeID 

•       store_and_fwd_flag 

•       PULocationID 

•       DOLocationID 

•       payment_type 

•       fare_amount (decimal data type) 

•       extra 

•       mta_tax 

•       tip_amount (decimal data type) 

•        tolls_amount (decimal data type) 

•       improvement_surcharge 

•       total_amount 

 

Tasks
VERY IMPORTANT:  for this question, you must first perform task a BEFORE performing task b, c, d, e and f. 

 

a)    [1 pts] Function load_data() to load data from Google Storage Bucket into Spark DataFrame

b)    [1.5 pts] Function exclude_no_pickuplocations() to exclude trips with no pickup locations (i.e., pickup location column is null or blank) in the original data from a. The Jupyter Notebook cell sequence automatically uses the original data from a. 

c)    [1.5 pts] Function exclude_no_tripdistance() to exclude trips with no distance (i.e., trip distance column is null or blank or zero) in the original data from a. The Jupyter Notebook cell sequence automatically uses the original data from a.

d)    [2 pts] Function include_fare_range() to include trips with fare from $20 (inclusively) to $60 (inclusively) in the original data from a. The Jupyter Notebook cell sequence automatically uses the original data from a.

e)    [2 pts] Function get_highest_tip() to identify the highest tip (rounded to 2 decimal places) in the original data. The Jupyter Notebook cell sequence automatically uses the original data from a.  

f)     [2 pts] Function get_total_toll() to calculate the total toll amount (rounded to 2 decimal places) in the original data. The Jupyter Notebook cell sequence automatically uses the original data from a. 

 


Q5 [10 points] Regression: Automobile price prediction, using Azure ML Studio 
Note: Create and use a free workspace instance on Azure ML Studio. Please use your Georgia Tech username (e.g., jdoe3) to login.   

Goal
Primary purpose of this question is to introduce you to Microsoft Azure Machine Learning Studio, familiarize you to its basic functionalities and typical machine learning workflows. Go through the “Automobile price prediction” tutorial and create/run ML experiments to complete the following tasks.  You will not incur any cost if you save your experiments on Azure till submission. Once you are sure about the results and have reported them, feel free to delete your experiments. 

Tasks
You will manually modify the given file q5_results.csv by adding to it the results from the following tasks

(e.g., using a plain text editor). 

•       DO NOT change the order of the questions.

•       Report the exact numerical values that you get in your output, and DO NOT round any of them.  

•       When manually entering a value into the csv file, append it immediately after a comma, so  there will be NO space between the comma and your value, and no trailing spaces or commas after your value.

•       Follow the tutorial and do not change values for L2 regularization. For parts b and c, please select the columns given in the tutorial.

 

a)    Update your GT username in the q5_results.csv file to replace gburdell3.

b)    [3 pts] Repeat the experiment described in the tutorial and report values of all metrics as mentioned in the ‘Evaluate Model’ section of the tutorial.  

c)    [3 pts] Repeat the experiment mentioned in part b with a different value of ‘Fraction of rows in the first output’ in the split module. Change the value to 0.8 from the originally set value, 0.75. Report corresponding values of the metrics.

d)    [4 pts] Run a new experiment — evaluate the model using 5-fold cross-validation (CV). Select parameters in the module ‘Partition and sample’ (Partition and Sample) in accordance with the figure below. Set the column name as “price” for CV.  Also, use 0 as a random seed. Report the values of Root Mean Squared Error (RMSE) and Coefficient of Determination for each of the five folds (1st fold corresponds to fold number 0 and so on). Do NOT round the results. Report exact values.

 

To summarize, for part d, you MUST exactly follow each step below to run the experiment:  

A.    Import the entire dataset (Automobile Price Data (Raw))

B.    Clean the missing data by dropping rows with missing values (select all columns in the dataset and do not "exclude the normalized losses" from the original tutorial).  Leave the maximum missing value ratio to 1.

C.   Partition and sample the data. (Note: do not use “Split Data”)

D.   Create a new model: Linear Regression (add the default Linear regression, i.e., do not change any values here)

E.    Finally, perform cross validation on the dataset. (Hint: use the price column here)

F.    Visualize/report the values.

 

 

Figure: Property Tab of Partition and Sample Module

 

Hint: For part 4, follow each of the outline steps carefully. This should result in 5 blocks in your final workflow (including the Automobile price data (Raw) block).