A Hybrid Explainable AI Approach for Enhanced Credit Risk Evaluation in Financial Services

Abstract

With the increasing complexity in financial services, machine learning models have to be sophisticated to enable a good estimation of credit risk. Yet, the lack of transparency and interpretability is common in traditional AI models, especially in financial decisions. In this paper, discuss the use of Explainable Artificial Intelligence (XAI) models to improve the credit risk assessment by providing predictive accuracy and transparency. The hybrid model presented is a combination of decision trees with the most recent XAI techniques, such as SHAP (SHAPley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations). The model was trained using publicly available credit data, on which there are variables of income, credit history, loan amount, and payment behavior. The findings reveal that the XAI model has an accuracy of 92, precision of 90, and recall of 88, which is higher than other traditional models. Moreover, the statistical model shows that the most important drivers of creditworthiness are payment history and income levels. The XAI model is very easy to use as it not only offers high performance but also builds trust and accountability in credit risk evaluation through providing actionable insights. As highlighted in this paper, explainable AI has the potential to revolutionize credit scoring systems, which would become more transparent and make the systems closer to regulatory standards. The further work will be dedicated to the implementation of other machine learning models and the use in the dynamic financial setting in real-time.