Enhancing Credit Risk Prediction through an Ensemble of Explainable Model

Abstract

Machine learning has been widely used in the field of credit scoring due to their excellent predictive performance, but opacity hinders the further application of more accurate complex machine learning models. We propose a credit score explainable framework that integrates multiple models, uses the XGBoost algorithm to predict credit scores, and then uses multiple explanation algorithms and K-means to enhance the accuracy and explainability of credit scores. The paper uses data sets from public websites to test model performance. The results show that the credit scoring model can simultaneously achieve the two goals of accurate prediction and stable explanation, making the credit scoring process easy to understand.

Keywords

Credit score / explanation / XGBoost / SHAP / LIME / K-means

References

[1] Apley D W, Zhu J. Visualizing the effects of predictor variables in black box supervised learning models. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2020, 82(4): 1059-1086.

[2] Ariza G, Miller J, Javier A, Antonio C, Maria-Jesus S. Explainability of a machine learning granting scoring model in peer-to-peer lending. IEEE Access, 2020, 8: 64873-64890.

[3] Barbaglia L, Manzan S, Tosetti E. Forecasting loan default in Europe with machine learning. Journal of Financial Econometrics, 2023, 21(2): 569-596.

[4] Basel Committee on Banking SupervisionInternational convergence of capital measurement and capital standards, 2006

[5] Bastos A, Sara M. Explainable models of credit losses. European Journal of Operational Research, 2022, 301: 386-94.

[6] Benhamou E, Ohana J J, Saltiel D, Guez B, Ohana S (2021). Explainable AI (XAI) models applied to planning in financial markets. SSRN Electronic Journal 10.

[7] Chen T, Guestrin C. Xgboost: A scalable tree boosting system. The 22nd ACM Sigkdd International Conference on Knowledge Discovery and Data Mining, 2016785-794

[8] Chen Y, Calabrese R, Martin-Barragan B. Interpretable machine learning for unbalanced credit scoring datasets. European Journal of Operational Research, 2024, 312(1): 357-372.

[9] Crone S F, Finlay S. Instance sampling in credit scoring: An empirical study of sample size and balancing. International Journal of Forecasting, 2012, 28(1): 224-238.

[10] De Lange P E, Melsom B, Vennerød C B, Westgaard S. Explainable AI for credit assessment in banks. Journal of Risk and Financial Management, 2022, 15(12): 556.

[11] Dumitrescu E, Hué S, Hurlin C, Tokpavi S. Machine learning for credit scoring: Improving logistic regression with non-linear decision-tree effects. European Journal of Operational Research, 2022, 297(3): 1178-1192.

[12] Friedman J H. Greedy function approximation: A gradient boosting machine. Annals of Statistics, 2001, 29(5): 1189-1232.

[13] Gramegna A, Paolo G. SHAP and LIME: An evaluation of discriminative power in credit risk. Frontiers in Artificial Intelligence, 2021, 4: 140.

[14] Gunnarsson B R, Vanden Broucke S, Baesens B, Óskarsdóttir M, Lemahieu W. Deep learning for credit scoring: Do or dont?. European Journal of Operational Research, 2021, 295(1): 292-305.

[15] Kozodoi N, Lessmann S, Papakonstantinou K, Gatsoulis Y, Baesens B. A multi-objective approach for profit-driven feature selection in credit scoring. Decision Support Systems, 2019, 120: 106-117.

[16] Liu Y, Huang F, Ma L, Zeng Q, Shi J. Credit scoring prediction leveraging interpretable ensemble learning. Journal of Forecasting, 2024, 43(2): 286-308.

[17] Lundberg S M, Lee S I (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems 30.

[18] Molnar C, Bischl B, Casalicchio G. iml: An R package for interpretable machine learning. Journal of Open Source Software, 2018, 3(26): 786.

[19] Papouskova M, Hajek P. Two-stage consumer credit risk modelling using heterogeneous ensemble learning. Decision Support Systems, 2019, 118: 33-45.

[20] Ribeiro M T, Singh S, Guestrin C. “Why should i trust you?” Explaining the predictions of any classifier. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 20161135-1144.

[21] Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors. Nature, 1986, 323: 533-536.

[22] Shapley L S. Stochastic games. Proceedings of the National Academy of Sciences, 1953, 39: 1095-1100.

[23] Sheng S H, Jin Z. Estimating credit risk parameters using ensemble learning methods: An empirical study on loss given default. Journal of Credit Risk, 2016, 12: 43-69

[24] Szepannek G. On the practical relevance of modern machine learning algorithms for credit scoring applications. WIAS Rep Ser, 2017, 29: 88-96

[25] Szepannek G (2019). How much can we see? A note on quantifying explainability of machine learning models. arXiv Preprint arXiv:1910.13376.

[26] Tanoue Y, Kawada A, Yamashita S. Forecasting loss given default of bank loans with multi-stage model. International Journal of Forecasting, 2017, 33(2): 513-522.

[27] Vapnik VThe Nature of Statistical Learning Theory, 2013Springer Science & Business Media

[28] Yao X, Crook J, Andreeva G. Support vector regression for loss given default modelling. European Journal of Operational Research, 2015, 240(2): 528-538.

RIGHTS & PERMISSIONS

Systems Engineering Society of China and Springer-Verlag GmbH Germany