A new research article “An explainable machine learning model for consumer credit scoring in Mexico," is published in Emerging Markets Review.

The study is a collaborative effort at the KDI School, co-authored by Professor Seohyun Lee, Professor Jaehyuk Park, and David Ugarte Chacon (PhD student), bringing together expertise across economics and computer science to develop an explainable ML framework for consumer credit defaults in an emerging market context. Empirically, the proposed XGBoost model outperforms both logistic and random forest models in predictive performance. 

Using SHAP to unbox the model, we find that financial attitudes and behaviors, among others, are major drivers of default risk.   We further evaluate the models through the SAFE framework (Sustainability, Accuracy, Fairness, and Explainability) and show that the XGBoost achieves strong predictive accuracy and fairness, while also exhibiting superior explainability compared to linear models.   

Our study highlights the potential of leveraging rich non-traditional and unstructured data in credit scoring for emerging countries with underdeveloped information infrastructure for the financial sector.

The article is freely accessible via the link below until January 30, 2026. 

https://authors.elsevier.com/a/1mFbq5a4WJeam8