120 likes | 282 Views
DSC 8330 Data Mining Project. Alex Kao Jack Peng Ben Hung. Score Card. Score Card Continue…. Score Card Continue…. KS Test. K-S Statistic For the test model: 22.10% For the validation subset: 21.82% For the whole dataset: 21.45% Cutoff Score: 630 Points
E N D
DSC 8330Data Mining Project Alex Kao Jack Peng Ben Hung
KS Test • K-S Statistic • For the test model: 22.10% • For the validation subset: 21.82% • For the whole dataset: 21.45% • Cutoff Score: 630 Points • About 75% of approved people are GOOD and 25% are BAD • Approved population has 69% of GOOD and 46% of BAD
KS Test Continue • Cutoff Score • For the test model: 630 Points • For the validation subset : 680 Points • For the whole dataset: 630 Points or 680 Points • If we select 680 Points as cutoff points • About 77.7% of approved people are GOOD and 22.3 % are BAD • Approved population has 51% of GOOD and 30% of BAD • Since the KS is very close and we want to approve more GOOD guys, we select 630 Points as our cutoff score
Monitor Report • Monitor and Track the performance of model regularly • Watch the GOOD/BAD distribution approved by the model after the GOOD/BAD rate becomes stable • Investigate the cause or consider revise/reconstruct the model when the demographic changes
conclusion • Cutoff point is 630, approval without conditions.
conclusion • Score between 600 and 630, conditional approval with higher interest rate • The model has higher approval rate, based on score 600, the rate is 72%.
Project Presentation Thank you