Explainable Machine Learning for Personal Credit Default Prediction

DOI

10.2991/978-94-6239-801-6_154How to use a DOI?

Keywords

credit default prediction; explainable machine learning; financial behavior features; threshold sensitivity; SHAP stability

Abstract

Predicting personal credit default is a vital part of managing financial risk for lenders. This paper proposes an explainable machine learning approach for predicting whether a credit card customer will default in the next month. The experiment employed the complete dataset of credit card clients from the UCI Machine Learning Repository, which includes 30,000 customer records. Three models are compared: Logistic Regression, Random Forest, and XGBoost. In order to make the study more meaningful than a simple model comparison, financial behavior features are constructed, including maximum repayment delay, average repayment status, credit utilization, and payment aggregates. The analysis further examines classification-threshold sensitivity and evaluates the stability of top-ranked explanatory features across five random seeds. The study revealed that XGBoost achieved the highest AUC score of 0.7822 and the highest recall score of 0.6436 after incorporating engineered features. Two of the most important and stable explanatory factors are the maximum repayment delay and the recent repayment status. The study argues that credit default models should be assessed not just for their predictive accuracy, but also for their decision thresholds and explanation quality.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Zixuan Qin
PY  - 2026
DA  - 2026/10/23
TI  - Explainable Machine Learning for Personal Credit Default Prediction
BT  - Proceedings of the 2026 2nd International Conference on Financial Innovation and Marketing Management (FIMM 2026)
PB  - Atlantis Press
SP  - 1466
EP  - 1473
SN  - 2352-5428
UR  - https://doi.org/10.2991/978-94-6239-801-6_154
DO  - 10.2991/978-94-6239-801-6_154
ID  - Qin2026
ER  -