Enhancing Credit Risk Decision-Making in Supply Chain Finance With Interpretable Machine Learning Model

The increasing complexity of supply chain finance poses significant challenges to effective credit risk assessment. Traditional black-box models often fail to provide insights into the factors driving credit risk, which is essential for stakeholders when making informed decisions. By conducting analysis of interpretable machine learning models, the study evaluated their performance in assessing credit risks. Specifically, we applied Extreme Gradient Boosting (XGBoost), Random Forest (RF), Least Squares Support Vector Machine (LSSVM) and Convolutional Neural Network (CNN) models for risk assessment. Our methodology included an ablation experiment along with utilizing Shapley Additive Explanation (SHAP) to elucidate the contribution and significance of specific risk factors. The results indicated that the asset-liability ratio, cash ratio, and quick ratio notably influence credit risk. This study clarified the applicability and limitations of various models, highlighting the superior performance and interpretability of XGBoost through the SHAP algorithm. Ultimately, the insights from this study provided valuable guidance for companies and financial institutions, fostering more sustainable allocation of financial resources.