search

ing-bank / skorecard

scikit-learn compatible tools for building credit risk acceptance models
https://ing-bank.github.io/skorecard/
MIT License
80 stars 22 forks source link

Fix inconsistencies with WoE #89

Open sbjelogr opened 1 year ago

sbjelogr commented 1 year ago

After the changes related to the WoEEncoder in #53 , there are still parts of the package that rely on the "old" calculation of WoE. Namely, this refers to the functions in 

scorecard.metrics.metrics

that are also used in the reporting functionality for IV, PSI etc.

Ideally, one would like that the WoE is calculated via the WOEEncoder everywhere, to avoid potential inconsistencies.

Update to the documentation might be needed as well. Check:

dlaprins commented 1 year ago

Related: the IV-score in bucket_table() has a negative sign, as it depends on the sign of WoE which was altered. This might be unintuitive and should probably be changed. (credit to Mara de Rosa for finding the issue)

josecaloca commented 3 months ago

I have noticed that the WoE calculations from bucket_table method are not consistent with the WoeEncoder(). I believe these calculations should be consistent as they might mislead analyses.

discrepancies WOE