Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{203597,
        author = {UDUGA SURYA KAMESWARI and Vassey Nagaraju and Gali Ratna sri and Kunchala Venkata Sai},
        title = {Explainable AI-Driven Hybrid TabTransformer and XGBoost Framework for Imbalanced Credit Risk Assessment},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {11195-11206},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=203597},
        abstract = {The problem of accurate credit scoring with explanations requires a simultaneous focus on identifying defaults and explaining decisions for those cases. This task poses significant challenges because real-world datasets suffer from severe class imbalance, where non-defaulters significantly outweigh defaulters. To address this, we present a novel hybrid stacked ensemble framework that combines a weighted-loss variation of TabTransformer with the XGBoost gradient boosting algorithm. Experiments were conducted on the Lending-Collection dataset (N = 32,416) with a class imbalance ratio of 3.57:1. We compared our proposed Hybrid Stack against baseline classifiers including Logistic Regression, Random Forest, XGBoost, and standard TabTransformer variants. While standalone XGBoost achieved the highest general accuracy (91.86%) and AUC-ROC (0.9446), the proposed Hybrid Stack proved superior for risk mitigation. The hybrid model achieved a competitive accuracy of 90.28% while maximizing the recall to 82.79% and the F2-score to 0.8117. This significantly improves the detection of minority class defaulters without requiring additional data preprocessing. Furthermore, SHAP analysis identified loan_int_rate, loan_percent_income, and debt_to_income_ratio as the primary drivers of model decisions, meeting regulatory requirements for explainability. Robustness was validated through 10-fold stratified cross-validation and McNemar significance tests, suggesting the framework is highly suitable for real-world financial risk system},
        keywords = {Credit scoring, class imbalance, explainable AI (XAI), SHAP, TabTransformer, XGBoost, hybrid stacked ensemble, financial risk assessment.},
        month = {May},
        }