New SHARC framework enhances explainability for ML risk models in finance

A new research paper introduces SHARC, an explainability framework designed for machine learning risk models used in regulatory capital estimation. SHARC addresses the 'black box' problem by applying SHapley Additive exPlanations (SHAP) to a Hybrid GPR-HS architecture, making its outputs auditable by supervisory bodies. The framework successfully links non-linear risk outputs to underlying scenario inputs and reveals that mean return components significantly influence capital levels under stress, offering implications for financial risk management. AI

IMPACT Enhances audibility and transparency of ML models in financial regulation, potentially accelerating adoption.

RANK_REASON The cluster contains a research paper detailing a new framework for machine learning interpretability in financial risk modeling. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• FRTB
• Hybrid GPR-HS
• SHAP

• paper
• policy

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →