International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Credit risk scoring systems are among the most crucial decision-making models developed by banks. These models have become mandatory in many jurisdictions because of regulatory requirements that promote risk-sensitive capital charge computations. However, legal requirements and the need for better customer relations now make credit-scoring evaluation systems necessary and increase demand in the market. Regulators require understandability of prediction models, banks are interested in interpretability to create customer relations and meet consumer-protection laws, and customers want to understand why they were rejected, especially in cases of marginal evaluation. Additionally, explainable artificial intelligence (xAI) models support the model-monitoring processes of banks and help the implementation of explainable credit assessment. The complex nature of the model-jungle makes it impossible for users and auditors to be aware of limitations, risks, and adequacy in risk management therefore clarity on how the models and systems become really interpretable is required.
Existing xAI credit scoring evaluation models, explainable AI (xAI) model families, and classes of post-hoc explanation techniques are examined to present a taxonomy of approaches. Based on the analysis, a set of operational and cognitive evaluation measures and compliance-oriented explainability tests are proposed. Finally, incorporation of these concepts into model deployment, xAI governance, and overall software life-cycle management in banks is outlined.

Keywords: Credit Risk Scoring Systems, Explainable Artificial Intelligence, Explainable Credit Assessment, Regulatory Compliance In Banking, Model Interpretability, Customer-Facing Credit Decisions, Consumer Protection Laws, Risk-Sensitive Capital Requirements, XAI Model Taxonomy, Post-Hoc Explanation Techniques, Credit Scoring Evaluation Frameworks, Model Transparency, Operational Explainability Metrics, Cognitive Evaluation Measures, Compliance-Oriented Explainability Tests, Model Monitoring And Validation, XAI Governance Frameworks, Banking Software Lifecycle Management, Risk Management Transparency, Interpretable Financial Models.

Downloads: | DOI: 10.17148/IJARCCE.2024.131267