★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★Open research. Global perspectives.
editor@theamericanjournals.comPublisher Home ↗
Articles OPEN ACCESS

Explainable AI driven Credit Risk Assessment for Basel III compliant Banking Risk Management

Aritra Pal
FIRST CITIZENS BANK Charlotte, North Carolina, USA
DOI Crossref DOI 10.37547/tajiir/Volume08Issue09-06
tajiir 2026
VOL. 8 / NO. 09 SEPTEMBER
FULL TEXT PP. 102-116
DOWNLOAD PDF

View all articles in this issue ↗

VOLUME 8
ISSUE 09
YEAR 2026
PAGES 102-116
RESEARCH OVERVIEW

Abstract

The effective identification of potential loan defaults is crucial for banking risk management, enabling informed lending decisions and financial stability. In this paper, a credit risk assessment framework that integrates both predictive modeling and risk interpretability is developed, which leverages Explainable AI (XAI) to enhance banking risk management to meet Basel III requirements. The proposed approach utilizes the probability outputs of the Random Forest (RF) and XGBoost (XGB) models which are combined using a Hybrid Model of Random Forest-XGBoost (RF-XGB). The Kaggle Credit Risk Dataset is preprocessed using duplicate remover, missing value imputer, outlier capping, categorical encoding, stratified data splitting and SMOTE data balancing. The performance of the Hybrid RF-XGBoost model is validated through experimental results with accuracy of 93.14%, precision of 94.19%, recall of 73.13%, F1 of 82.33%, ROC-AUC of 93.28%, and PR-AUC of 88.48%, for the default prediction task. In order to increase transparency of the model, explanations of individual predictions and influential credit-risk factors are provided by means of SHAP and LIME. The predicted Probability of Default (PD) is then translated to five risk categories (Very Low Risk through Very High Risk) and combined with Loss Given Default (LGD), Exposure at Default (EAD), Expected Loss (EL), risk-weighted assets (RWA) and capital requirements. Additionally, a stress testing process is conducted to assess for possible rise in credit losses during adverse conditions. The framework therefore offers a holistic solution which integrates predictive accuracy with explanations, borrower-level risk stratification and Basel III oriented risk assessment. The findings suggest that combining XAI with ensemble credit risk models can boost the transparency and usability of automated banking credit decisions.

RESEARCH TOPICS

Keywords

Explainable Artificial Intelligence (XAI) Credit Risk Assessment Basel III Probability of Default SHAP LIME Credit Risk Management

Downloads

Download data is not yet available.
SOURCES & SCHOLARSHIP

References

  1. P. E. de Lange, B. Melsom, C. B. Vennerød, and S. Westgaard, “Explainable AI for Credit Assessment in Banks,” J. Risk Financ. Manag., vol. 15, no. 12, p. 556, Nov. 2022, doi: 10.3390/jrfm15120556.
  2. [V. P. Rambabu, J. Sreerama, and J. T. S. Singh, “AI-Driven Data Integration: Enhancing Risk Assessment in the Insurance Industry,” Aust. J. Mach. Learn. Res. Appl., vol. 2, no. 2, pp. 130–179, July, 2022.
  3. D. O. Oyeyemi, O. O. Okosieme, O. Idowu-Kunlere, E. A. Julius, and I. P. Nwinyi, “Explainable AI For Credit Risk Assessment: Integrating Machine Learning With Business Analytics,” IOSR J. Econ. Financ., vol. 16, no. 5, pp. 63–72, 2025, doi: 10.9790/5933-1605016372.
  4. M. Rambauli, T. Ravele, and C. Sigauke, “Application of Explainable AI and Uncertainty Quantification in Credit Risk Assessment,” Risks, vol. 14, no. 7, p. 152, Jul. 2026, doi: 10.3390/risks14070152.
  5. N. Bussmann, P. Giudici, D. Marinelli, and J. Papenbrock, “Explainable Machine Learning in Credit Risk Management,” Comput. Econ., vol. 57, no. 1, pp. 203–216, Jan. 2021, doi: 10.1007/s10614-020-10042-0.
  6. P. Chawande, “Model Risk Governance for AI-Based Compliance Systems in Investment Banking,” Int. J. Multidiscip. Res. Growth Eval., vol. 6, no. 3, pp. 2027–2035, 2025, doi: 10.54660/.IJMRGE.2025.6.3.2027-2035.
  7. S. Malave, B. Khemani, H. Patil, S. Nandurkar, O. Nandurkar, and A. Nayak, “Transforming financial documents into credit decisions using explainable artificial intelligence and optical character recognition,” MethodsX, vol. 16, p. 103962, Jun. 2026, doi: 10.1016/j.mex.2026.103962.
  8. M. Indira, R. M. Basha, P. Sivamma, M. S. Babu, and S. H. N. Ketan, “Attention-Balanced Variational Autoencoder Framework for Credit Card Fraud Detection under Extreme Class Imbalance,” in 2026 7th International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV), IEEE, May 2026, pp. 1060–1067. doi: 10.1109/ICICV68925.2026.11554839.
  9. S. Kakkar, “Explainable AI Models for Credit Risk Scoring in Banking: Balancing Accuracy and Regulatory Transparency,” Int. J. Financ. DATA Sci., vol. 3, no. 2, pp. 1–6, Aug. 2025, doi: 10.34218/IJFDS_03_02_001.
  10. M. J. Lim, P. W. Alexander, V. M. Karyanto, T. C. Christie, and T. Oktavia, “Detecting Credit Card Fraud using Random Forest and CRISP-DM: A Machine Learning Approach,” in 2025 International Conference on ICT for Smart Society (ICISS), IEEE, Sep. 2025, pp. 1–5. doi: 10.1109/ICISS66954.2025.11389400.
  11. S. Chauhan, K. S. Gill, R. Chauhan, and H. S. Pokhariya, “Smart Solutions for Utilizing Advanced Machine Learning for Robust Credit Card Fraud Detection,” in 2024 First International Conference on Electronics, Communication and Signal Processing (ICECSP), IEEE, Aug. 2024, pp. 1–6. doi: 10.1109/ICECSP61809.2024.10698053.
  12. Z. Xu, J. Cheng, L. Cheng, X. Xu, and M. Bilal, “MSEs Credit Risk Assessment Model Based on Federated Learning and Feature Selection,” Comput. Mater. Contin., vol. 75, no. 3, pp. 5573–5595, 2023, doi: 10.32604/cmc.2023.037287.
  13. L. Tse, “Credit Risk Dataset,” Kaggle.
  14. Shreya and H. Pathak, “Explainable Artificial Intelligence Credit Risk Assessment using Machine Learning,” arXiv Comput. Sci., Jun. 2025, doi: 10.48550/arXiv.2506.19383.
  15. K. N. Islam, “An Explainable Machine Learning Model for Credit Risk Prediction: Evidence from Commercial Banks in Bangladesh,” Am. J. Ind. Bus. Manag., vol. 16, no. 07, pp. 647–673, 2026, doi: 10.4236/ajibm.2026.167034.