lr-modeling

What problem does it solve?

This Skill solves the problem of producing an explainable binary risk scorecard when you need a transparent model for regulated or business-auditable decisioning.

Core Features & Use Cases

• WoE optimal binning + Logistic Regression: Converts raw features into WoE-encoded bins, trains a Logistic Regression model, and derives a standard scorecard mapping.
• Regulatory-friendly artifacts: Outputs coefficient tables, WoE-to-score contributions, a deployable scorecard JSON (bin-to-score mapping), and a Markdown modeling report with AUC/KS/BCR and stability sections.
• Uses standard data splits and evaluations: Reuses the platform’s three-stage split (train/val/oot), AUC/KS evaluation, and stability analysis hooks; designed for production workflow consistency.

Quick Start

Use the lr-modeling skill with your dataset file, specifying the target column and optional time column, to generate the WoE+LR scorecard artifacts and evaluation report in an output directory.