In recent years, banking systems have faced a significant increase in both the volume and complexity of customer financial data. As a result, traditional credit scoring methods based on linear assumptions and limited indicators have become less effective. Machine learning techniques have therefore emerged as powerful tools for credit risk assessment due to their ability to capture complex nonlinear relationships in large datasets. However, a major limitation of these models is their lack of interpretability, which restricts their practical use in regulated banking environments.
This study proposes an explainable credit scoring framework by integrating machine learning models with SHAP. Behavioral features such as transaction patterns, debt levels, repayment history, and other customer-related indicators are used to build predictive models. Algorithms including Random Forest, Gradient Boosting Machines, and Support Vector Machines are applied to evaluate credit risk prediction performance. SHAP is then used to interpret model outputs and quantify the contribution of each feature.The results show that repayment history, debt-to-income ratio, and transaction behavior are the most influential factors in determining credit scores. The integration of SHAP with machine learning not only maintains high predictive accuracy but also significantly improves model transparency and interpretability. Overall, the proposed approach enhances trust, fairness, and accountability in credit scoring systems and provides a strong foundation for future research in explainable financial analytics.