Question 1
Imagine you were the manager of the Bank ABC.
You would like to find out what financial products/services are more likely to
be bought together by your customers, so that you and your team can better
design recommendations and advertising campaigns for your financial services in
the coming year. After consulting with a senior business analytics specialist,
you decide to conduct a market basket analysis.
The IT department has done the data
preparation as per your requirement. A BANK data set that contains service
information of thousands of customers is ready for analysis. There are three
variables in the dataset, as shown in the table below.
Field |
Type |
Description |
ACCOUNT |
Nominal |
Account number |
SERVICE |
Nominal |
Type of product service |
VISIT |
Ordinal |
Order of product purchase |
The 13 products are represented in the data
set using the following abbreviations:
ATM
automated teller machine debit card
AUTO
automobile installment loan
CCRD credit card
CD certificate of deposit
CKCRD
check/debit card
CKING checking account
HMEQLC
home equity line of credit
IRA individual retirement account
MMDA
money market deposit account
MTG mortgage
PLOAN
personal/consumer installment loan
SVG saving account
TRUST personal trust account
(a)
Data exploration: what is the format of the data? How many individual accounts
are there in this dataset? What are the top three (3) product/services that the
bank customers have bought? Show how you get the results. Evaluate the
suitability of using Association Rule Mining for this problem. (15 marks)
(b)
Construct an Apriori model on the dataset using IBM SPSS Modeler. The model
details and interpretation of results should include the following:
(i)
Report the “Fields” setting of the Apriori node. Give a screenshot of the
setting. (ii) Set the Minimum Support = 10%, Minimum Confidence = 60%, Maximum
number of antecedents = 5. Report the
number of rules generated, and give a screenshot of the rules. (iii) Observe
the top 10 rules that have the highest Confidence values; what is the key
pattern you see when interpreting these rules in general (no need to list out
all the 10 rules)? (iv) Report two (2) other interesting rules and their
implications. (25 marks)
(c)
Distinguish Sequence Pattern Mining from ARM. Evaluate whether Sequence Pattern
Mining can be used to study this dataset.
(10 marks)
Question 2
One
common issue with most traditional Association Rule Mining (ARM) algorithms
(e.g., Apriori, CARMA) is their inability to mine numerical data without first
converting them into categorical ones. Write a research essay to discuss this
issue and critically review at least one important research article that
attempts to address this issue. The review should include the technique
description, advantage and possible limitation discussion of the proposed
method in the research article. Keep the essay length to two pages. (40 marks)
Get Free Quote!
439 Experts Online