Explainable AI for Loan Approval Decisions: Integrating Contextual SHAP with Dynamic Temporal Analysis

This paper addresses the challenge of achieving both high predictive accuracy and interpretability in AI-based loan approval systems. To solve problems connected with black boxes, a novel approach was introduced C-SHAP-DTA, which includes the elements of Dynamic Temporal Analysis (DTA) and Contextual SHAP (C-SHAP). C-SHAP-DTA understands temporal financial behavior through the means of anomaly detection with autoencoders and provides readable explanations on account of clustering techniques in the field of context. It turns out that the proposed approach is more successful than the alternatives, as it demonstrates 87% accuracy, 89% precision, and the trust score of 4.1 with strong correlation of time. Such results highlight the effectiveness of integrating two important components into a system for evaluating credit risks time dynamics and contextual changes. In practice, the suggested approach can be used not only to make reliable decisions but also improve transparency, personalization, and fairness of operations. Moreover, it helps institutions to build trust, compliance, and increase their overall reliability. As far as performance is concerned, the model demonstrates both accuracy and interpretability by providing time and applicant-specific explanation.