Data Mining Project Presentation

1
Data Mining Project Presentation

• Chris Hickman
• Nathaniel Owen
• April 24, 2006

2
Introduction

• To predict from past credit data which customers
  will be bad customers and which customers will
  be good customers.
• Bad customers those who default on the loan
• Good customers those who pay off the loan in a
  matter suitable to the credit loan bureau

3
K-S Test - Training Results

• Maximum Difference 0.22

4
K-S Test - Validation Results

• Maximum Difference 0.210

5
K-S Test Combined Data Results

• Maximum Difference 0.215

6
Scorecard
7
Scorecard (continued)
8
Scorecard (continued)
9
Customer Profile

• Above Age 45
• High Down Payment (above 1500 dollars)
• Conservative with their Credit History

10
Monitoring Report Score Distribution

• Maximum Observed Difference Computed Quarterly
• Minimum Required Difference at 95 Confidence
  Level 4.27

11
Monitoring Report Characteristic Distribution
12
Monitoring Report Characteristic Distribution
(cont.)
13
Monitoring Report Characteristic Distribution
(cont.)
14
Recommendation

• At a score of 625 there is a high percentage of
  Good customers and a manageable number of Bad
  customers.
• Any customer falling below 625 should be
  considered risky for financing and offered the
  appropriate interest rate.
• Any score above 625 should be considered less
  risky and given an interest offer to coincide
  with the score

15
Questions?????