BIA 660 Web Analytics - Midterm Akshta Chougule Hao Han Di Huo Xi Lu Laura Sills Bank Of America.

BIA 660 Web Analytics - Midterm

Akshta ChouguleHao HanDi HuoXi LuLaura Sills

Bank Of America

Business Problem

Customer Strategy: grow base by forming life-long banking relationships with young adults

Current Account Demographics Report Shows● fewer new student accounts● increase in cancellation of accounts by

the young adult demographic

Impact: Losing market share to other banks

Business Questions

● What is Bank of America’s reputation with this age group - do they like Bank of America or not?

● How does Bank of America compare to other banks?

● Are customers in this demographic group unhappy with the bank’s services?

● Are there any banking products which customers in this group want not offered by Bank of America?

Source of Information

Online social media sites are a good source for comments from this age group

YouTube Statistics

●More than 1 billion unique users monthly

● Nielsen ratings show that YouTube reaches more US adults ages 18-34 than any other cable network

http://www.youtube.com/yt/press/statistics.html

Demographics of Reddit

http://www.theatlantic.com/technology/archive/2013/07/reddit-demographics-in-one-chart/277513/

What do People Think About Banks?

Topic Reddit YouTube Twitter

mortgage 5% 6% 30%

loan 5% 13% 0%

fraud 6% 7% 0%

insurance 1% 2% 0%

branch 3% 1% 0%

hours 2% 1% 0%

account 19% 16% 20%

overdraft 8% 1% 0%

bailout 1% 6% 0%

fee 18% 11% 20%

customer 13% 8% 0%

representative / teller 7% 18% 20%

[credit] union 10% 7% 10%

computer 1% 1% 0%

CEO 2% 2% 0%

Data Gathering and Validation

Use Python to obtain comments from web

● Crawling Reddit

● API for Twitter

● API for YouTube

Data Cleansing and Exploration

● Delete incomplete comments, extra whitespace, and punctuation, stopwords

● Explore data using Python to analyze the frequency of words in the comments in order to identify “key words” related to banking

● Word scan confirmed the key words

Gathering data from Twitter ● Technique: twitter API● Amount of tweets:

BOA -- 125KB

Citibank-- 104 KB

Chase -- 100 KB

● Timestamp: 1 week ● Type of Data:

Tweet text

Tweet created_at

Geocode

Data Processing

● Two libraries: positive & negative

● Score each tweet

Tweets by Location

Data Processing

● Summary for BOA tweets:

● Good or bad?

Min. 1st Qu. Median Mean 3rd Qu. Max.

-0.20000 -0.04348 0.00000 -0.01176 0.02857 0.20000

Competitor Analysis

Distribution for tweets’ score

Mean:

BOA: -0.01176

Citi bank: -0.0006146

Chase: -0.00731

Two Sample T-test

Null hypothesis: true difference in means is equal to 0Alpha=0.1

● BOA and Citi bank:

p-value = 0.0009004 < 0.1● Citi bank and Chase:

p-value = 0.06971 < 0.1● BOA and Chase

p-value = 0.2289 > 0.1

Gathering data from YouTube

● Techniques: BeautifulSoup

g.data

● Amount for general analysis: 3097

Topic Reddit YouTube Twitter

mortgage 5% 6% 30%

loan 5% 13% 0%

fraud 6% 7% 0%

insurance 1% 2% 0%

branch 3% 1% 0%

hours 2% 1% 0%

account 19% 16% 20%

overdraft 8% 1% 0%

bailout 1% 6% 0%

fee 18% 11% 20%

customer 13% 8% 0%

representative / teller 7% 18% 20%

[credit] union 10% 7% 10%

computer 1% 1% 0%

CEO 2% 2% 0%

YouTube data for each category

● Training data: 600

● Loan: 2430

● Account: 2700

● Service: 520

Naive Bayes Classification Algorithm

A naive Bayes classifier assumes that the presence or absence of a particular feature is unrelated to the presence or absence of any other feature, given the class variable。

Naive Bayes Classification Algorithm

Splitting the dataset into training and test data

(Manual rating of comments)

● Training (400)

● Testing (200)

● Predicting (5700)

Primary Categories of Customer Complaints

Accuracy of Classification

● Mortgage: 64.5%

● Accounts: 58.7%

● Service: 68.4%

Mortgage

Account

Service

Thank you!