← Back to Projects Finance 
Credit Risk Scoring Model
Interpretable ML model for loan approval using LightGBM with SHAP/LIME explainability, Fairlearn bias testing, traditional WoE/IV scorecard, and regulatory compliance.
Overview
Interpretable credit risk scoring system with built-in fairness testing and regulatory compliance reporting.
Architecture
- LightGBM with Optuna hyperparameter optimization
- SHAP and LIME for model interpretability
- Fairlearn for bias detection and mitigation
- Traditional WoE/IV scorecard for comparison
- FastAPI serving endpoint
Key Features
- Dual-model approach: ML + traditional scorecard
- Automated fairness auditing across protected groups
- Regulatory compliance report generation
- Feature contribution explanations per prediction