An Interpretable Temporal Domain-adaptive Survival Boosting Framework for Small-business Loan Charge-off Risk Assessment: Evidence from 2020-2026 SBA 7(a) loans
Abstract
Small-business credit-risk assessment is complicated by class imbalance, right censoring, and temporal distribution shift. This study
develops an interpretable temporal domain-adaptive survival boosting (TDASB) framework for assessing 24-month charge-off risk in U.S.
Small Business Administration (SBA) 7(a) loans. Using public SBA 7(a) FOIA data through 31 March 2026, 304, 570 originated loans with 5,
793 observed charge-offs are retained. TDASB combines censoring-aware cohort construction, temporal density-ratio weighting, and discretetime survival boosting. An out-of-time experiment trains on FY2020-FY2022, validates on FY2023, and tests on mature FY2024 loans.
TDASB achieves the highest PR-AUC of 0.449 and captures 75.2% of observed charge-offs within the highest-risk 10% of the test portfolio,
indicating improved high-risk concentration while retaining managerial interpretability.
develops an interpretable temporal domain-adaptive survival boosting (TDASB) framework for assessing 24-month charge-off risk in U.S.
Small Business Administration (SBA) 7(a) loans. Using public SBA 7(a) FOIA data through 31 March 2026, 304, 570 originated loans with 5,
793 observed charge-offs are retained. TDASB combines censoring-aware cohort construction, temporal density-ratio weighting, and discretetime survival boosting. An out-of-time experiment trains on FY2020-FY2022, validates on FY2023, and tests on mature FY2024 loans.
TDASB achieves the highest PR-AUC of 0.449 and captures 75.2% of observed charge-offs within the highest-risk 10% of the test portfolio,
indicating improved high-risk concentration while retaining managerial interpretability.
Keywords
Small-business lending; Credit risk; Charge-off prediction; Domain adaptation; Survival learning; Gradient boosting; Explainable AI
Full Text:
PDFReferences
[1] Bellotti, T., & Crook, J. (2009). Support vector machines for credit scoring and discovery of significant features. Expert Systems with
Applications, 36(2), 3302-3308. https://doi.org/10.1016/j.eswa.2008.01.005
[2] Breiman, L. (2001). Random forests. Machine Learning, 45, 5-32. https://doi.org/10.1023/A:1010933404324
[3] Cox, D. R. (1972). Regression models and life-tables. Journal of the Royal Statistical Society: Series B, 34(2), 187-220. https://doi.
org/10.1111/j.2517-6161.1972.tb00899.x
[4] Csurka, G. (Ed.). (2017). Domain adaptation in computer vision applications. Springer. https://doi.org/10.1007/978-3-319-58347-1
DOI: http://dx.doi.org/10.70711/frim.v4i9.10223
Refbacks
- There are currently no refbacks.
