International Journal of Scientific Research and Engineering Development

Credit risk assessment increasingly relies on high-performance black-box ML models such as XGBoost and deep neural networks. While these models deliver superior predictive accuracy, their opacity conflicts with regulatory frameworks comprising GDPR‘s right-to-explanation and fair lending laws. This paper applies and compares three post-hoc XAI techniques; SHAP, LIME, and counterfactual clarifications; on various and several black-box models trained on a real-world credit dataset (for example, German Credit, HELOC, LendingClub). We evaluate each strategy on faithfulness, stability, computational cost, and human interpretability. Results show that SHAP offers the most globally uniform and unchanging clarifications, while counterfactuals are most actionable for individual loan applicants. We further discuss implications for regulatory adherence and propose a decision-support framework for deploying explainable credit scoring in practice.

Charvak Dharmaraj Chavare, Chaitanya Ramesh Darekar,"Towards Interpretable Credit Risk Assessment: A Comparative Study of XAI Techniques with Regulatory Compliance" International Journal of Scientific Research and Engineering Development, V9(3): Page(283-306) May-June 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.