A Validation-Weighted Explainable Ensemble for Consumer Credit Default Prediction: Balancing Predictive Performance and Regulatory Interpretability

DOI

10.2991/978-94-6239-787-3_49How to use a DOI?

Keywords

Consumer Credit Risk; Credit Default Prediction; Ensemble Learning; Explainable AI; SHAP; FinTech Governance

Abstract

The growth of digital and online consumer lending has intensified the need for accurate yet transparent default-prediction models. Black-box machine-learning models often achieve strong discrimination but are difficult to deploy under financial-regulatory regimes that demand explainability. This paper proposes an explainable ensemble-learning framework that couples five heterogeneous base learners—logistic regression, random forest, extremely randomised trees, gradient boosting and histogram-based gradient boosting—through a convex meta-combination whose weights are selected to maximise the area under the ROC curve on a held-out validation set, and integrates a model-agnostic Shapley-value (SHAP) explanation layer. Synthetic minority over-sampling (SMOTE) mitigates class imbalance, and the conclusions are shown to be robust to this choice. On the public ‘default of credit card clients’ benchmark (30,000 records, 22.1% default rate), the proposed ensemble attains the highest AUC (0.776) and Kolmogorov–Smirnov statistic (0.419); the margin over the strongest single learner is small but statistically significant (DeLong p < 0.05) and stable across five repeated runs (mean AUC 0.781 ± 0.007), and the validation-based weighting effectively reduces to a two-learner combination of the bagging models. The SHAP analysis identifies the recent repayment-status history and the credit limit as the dominant default drivers and yields per-applicant explanations for adverse-action reporting, and a preliminary subgroup analysis shows only a small fairness gap across applicant sex. The results indicate that a carefully designed ensemble can improve ranking performance without sacrificing the transparency required for compliant credit decision-making.

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  - Kelin Qian
AU  - Zhanpeng Wang
AU  - Chunjiang Li
PY  - 2026
DA  - 2026/09/29
TI  - A Validation-Weighted Explainable Ensemble for Consumer Credit Default Prediction: Balancing Predictive Performance and Regulatory Interpretability
BT  - Proceedings of the 2026 4th International Conference on Management Innovation and Economy Development (MIED 2026)
PB  - Atlantis Press
SP  - 495
EP  - 504
SN  - 2352-5428
UR  - https://doi.org/10.2991/978-94-6239-787-3_49
DO  - 10.2991/978-94-6239-787-3_49
ID  - Qian2026
ER  -