This assignment is intended to get you thinking about association rules and simple data mining techniques in a way that is 1) applicable to formation systems, 2) applicable to a real-world decision, and 3) interesting!

Assignment:

For this assignment, pretend that you are a newly hired loan officer at a bank. Your new job is to approve or deny loan applications. You must approve all or none of a loan- you cannot approve part of it. Since you are new, you dont have much experience. However, your boss did give you a dataset

credit_risk_assn2_workbook_final.xlsx

The dataset included information about 425 clients that the bank has done business with in the past. The dataset tells you some facts about the clients in the form of several independent variables (demographics, bank account info, past loan outcomes, etc.), and it also tell you whether the bank thinks these individuals are a good credit risk (indicated by the dependent variable credit standing). What you would like to do is learn from these examples. In the very near future, you will be asked to process a loan application. Of course, since you are new, you will not have the final say- your boss or your bosss boss will, depending on the amount of the loan. Your boss will be watching you to see if you make stupid loans to bad clients, and also if you miss opportunities to make prudent loans to good clients.

Fortunately, you paid attention in college, and you seem to remember a technique for finding rules that might help you make decisions. Oh, yeah- association rules was one technique you seem to remember. But, thats all you remember. So, to get brushed up, you watch the two lectures your teacher posted about association rules.

The lectures that accompany this project can be found at:

Part 1

http://www.youtube.com/watch?v=v0vdP5fqxF4&feature=youtu.be

Part 2

http://www.youtube.com/watch?v=DaO4rM0Ftv0&feature=youtu.be

Finally, you think you might want to try to come up with some association rules. So, the assignment is as follows:

Come up with between at 15-30 association rules like the ones you see on the right-hand side of the spreadsheet. Your rules can be on the right-hand side of the same worksheet , or you may want to list them on a separate worksheet (please keep them in the same workbook/file).

For each association rule, compute the support and confidence of each rule.

What you want to do is find a set of rules that will ID both bad and good loan risks. So, ID some great rules, and then write a short memo for your boss. The memo should be at least 2 paragraphs but less than 2 pages. Assume he knows nothing about association rules. The goal is to explain why the rules you have discovered justify not firing you this Friday. (Basically- explain what you have discovered and why it is significant). DO NOT give a walk-through for each rule you discover. Rather, you might want to choose 2-3 especially informative rules and briefly discuss them as examples, and assume your boss will get the point.

DO NOT USE MACROS in the file.

DO NOT save your data to an external file or import data from another file! Your work should be self-contained within your Excel workbook.

Telephone

Foreign

Age

Credit Standing

Yes

Yes

23

Good

Yes

Yes

32

Bad

Rule

Count

Support

Confidence

No

Yes

38

Bad

Single

233

Yes

Yes

36

Bad

Single => good

130

0.305882

0.55794

No

Yes

31

Good

Single => bad

103

0.242353

0.44206

Yes

No

25

Good

Divorced

156

Yes

Yes

26

Good

Divorced => good

65

0.152941

0.416667

Yes

Yes

27

Good

Divorced => bad

91

0.214118

0.583333

Yes

Yes

25

Bad

Married

36

No

Yes

43

Bad

Married => good

19

0.044706

0.527778

No

Yes

32

Bad

Married => bad

17

0.04

0.472222

Yes

No

34

Good

1

No

Yes

26

Good

Yes

Yes

44

Bad

Single and Savings High)

9

Yes

Yes

46

Good

(Single and Savings High)=>good

7

0.016471

0.777778

No

Yes

39

Good

No

Yes

25

Bad

(Single and Savings MedHigh)

16

Yes

Yes

31

Good

(Single and Savings Med) =>good

10

0.023529

0.625

No

Yes

47

Good

Yes

Yes

23

Bad

Single and (Savings High OR MedHigh)

25

Yes

Yes

22

Bad

Single and (Savings High OR MedHigh)=>good

17

0.04

0.68

Yes

Yes

26

Bad

Yes

Yes

19

Bad

No

Yes

27

Bad

Yes

No

39

Good

Yes

Yes

26

Good

No

Yes

50

Bad

No

Yes

34

Good

Yes

Yes

23

Good

Yes

Yes

23

Good

No

Yes

46

Bad

Yes

Yes

35

Good

Yes

Yes

28

Good

No

Yes

25

Good

Yes

Yes

36

Bad

Yes

Yes

41

Good

No

Yes

54

Bad

Yes

Yes

43

Good

No

Yes

33

Bad

Yes

Yes

34

Bad

No

Yes

39

Good

No

Yes

34

Good

Yes

Yes

30

Good