Construction of a Supply Chain Credit Risk Evaluation Model for Manufacturing Enterprises Using XGBoost

Abstract

In the context of increasingly globalized and complex supply chain networks, manufacturing enterprises are exposed to growing credit risk challenges, particularly in the realm of supply chain finance. Traditional credit risk evaluation methods—such as logistic regression or manual scoring—often fail to capture nonlinear interactions and hidden patterns in high-dimensional enterprise data, leading to suboptimal risk classification. To address these limitations, this study proposes a data-driven credit risk evaluation model based on the eXtreme Gradient Boosting (XGBoost) algorithm. The model integrates multidimensional features, including financial indicators, transactional behavior, and credit-related attributes of upstream and downstream supply chain partners. By leveraging XGBoost's powerful capabilities in feature selection, regularization, and handling missing values, the model provides more accurate and interpretable credit risk predictions. Comprehensive experiments on real-world manufacturing enterprise datasets demonstrate that the proposed model significantly outperforms benchmark models—such as logistic regression, random forest, and support vector machine—in terms of AUC, precision, recall, and F1-score. Furthermore, feature importance analysis offers managerial insights into the key drivers of supply chain credit risk. This research contributes to both academic understanding and practical application by offering a scalable and intelligent approach to risk assessment, supporting more informed decision-making for supply chain managers, financial institutions, and policy makers.