lr-modeling

To build an explainable credit risk scorecard, this Skill converts input features into WoE optimal bins and fits a Logistic Regression model, outputting transparent bin-to-score mappings for regulated, auditable decisioning.

Binary risk modeling outputs include standard AUC and KS evaluation metrics, alongside OOT validation and stability monitoring reports, ensuring your scorecard meets production workflow consistency requirements.

For WoE+LR scorecard development, you need a pre-built dataset with a 0/1 target column, an optional time column for OOT validation, and a three-stage splitting configuration across train, validation, and out-of-time sets.

Yes, the logistic regression risk modeling workflow leverages scikit-learn for the LR model and includes xgboost alongside optbinning and scipy to support WoE optimal binning and statistical computations.

Creating a white-box scorecard generates coefficient tables, WoE-to-score contributions, a deployable scorecard JSON with bin-to-score mapping, and a Markdown modeling report detailing AUC/KS/BCR and stability sections.

Choose WoE+LR over other risk modeling approaches when you need a transparent, white-box model for regulated or business-auditable decisioning, as it provides clear WoE-to-score contributions rather than opaque predictions.