This document analyzes three years of sales data for an online retail company using tools like Microsoft Azure, Hadoop, Hive, and Spark. It finds that Spark queries were faster than Hive queries. It also uses the Apriori algorithm to determine association rules between product subcategories, containers, and shipping modes. Graphs show trends in sales by province and customer segment. However, no correlation was found between holidays and increased sales. The analysis aims to help the company increase yearly revenues by predicting demand and optimizing pricing.

1. Dashboard 1
2
1
Big Data Science
Online Retail Store Analysis
Submitted to : Dr Jongwook Woo
24th Annual Student Symposium,CSULA
Submitted By : Rajeev Singh , Manvi Chandra
Richa Kankarej
California State University, Los Angeles

2. 2

3. 3

4. 4

5. 5

6. 6

7. 7

8. 8

9. 9

10. 10

11. 11

12. 12

13. 13

14. 14

15. 15

16. 16

17. 17

18. 18

19. 19

20. Introduction…
 Industry : Online Retail (Canadian Company)
 In this foundational white paper, we used
Microsoft Azure, Hadoop, Hive, Spark and Apriori
Algorithm to model and analyze bid data for
GroupX
 Requirements : Analyzed 3 years of historical
data for peak sales and high net revenue
generating customers
Submissions
20

21. Go Big OR Go Home….
21
Retail Industry – It (Online & Offline) is huge of approx.
8 Trillion USD
Extrapolation – From weather patterns, search/browsing
trends, social networks, industry forecasts, existing
customer records
Predict – Instore sales, predict product trends, forecast
demand, pinpoint customer, optimize pricing and
promotions
Leverage to Retailers – Fewer sotckouts, higher visit to
buy ratio, better anticipation and response to market
shifts.
Retail Oligopoly – A market structure where only few
firms dominate.
Go Big …

22. Flow Diagram
22

23. Hive Query for Valued Customer
23

24. Hive Query for Sale Trend
24

25. Spark Product SubCatg v/s Sum(sales)
25

26. Spark Province v/s sum(sales)
26

27. Spark Customer Segment v/s Sum(Sales)
27

28. 28
Query In Hive In Spark
Sales in Province 56.01 sec 16.2 sec
Holiday Trend 4.21 sec 2.4 sec
Product SubCatg and Sum(Sales) 62.8 sec 8.7 sec
Customer Segment and Sum(Sales) 74.14 sec 5.8 sec
Time comparison of Hive and Spark Query

29. Apriori Algorithm / Predictive Analytics
 For frequent item set mining over transactional databases
 To determine association rules which highlight general trends in
the database
 This has applications in domains such as market basket analysis
 SAP Predictive Analytics
29

30. Rules between SubCatg, Container, Ship Mode
30
Rules Confidence
{SubCatg=Binders and Binder Accessories} =>
{Container=Small Box}
1
{SubCatg=Binders and Binder Accessories} =>
{Ship_Mode=Regular Air}
0.88
{Container=Small Pack} => {Ship_Mode=Regular Air} 0.87
{Container=Wrap Bag} => {Ship_Mode=Regular Air} 0.88
{SubCatg=Paper} => {Ship_Mode=Regular Air} 0.87

31. 31
Representation of Rules

32. Holiday Trend graph
32

33. Conclusion/Learnings
• No co-relation in Holiday season and an increase in sales for GroupX
• Spark is faster than Hive
• What steps can GroupX take to increase its YoY(Year on Year)
revenues
33

34. References
 References
 What is Hive?
 http://www-01.ibm.com/software/data/infosphere/hadoop/hive/.
 Introduction to Hadoop in HDInsight: Big-data analysis and processing in the
cloud. https://azure.microsoft.com/en-us/documentation/articles/hdinsight-
hadoop-introductin
34

35. Dashboard 1
2
35
Thank You
Q & A