Comparative analysis of class imbalance reduction methods in building machine learning models in the financial sector

A.F. Konstantinov, L.P. Dyakonova

---

Abstract: Borrower default prediction is a pressing issue that underlies the financial stability of credit institutions.
Aim. This study is to develop and evaluate an integrated borrower default prediction method.

Materials and methods. The study was conducted by simulating the integrated borrower default prediction method, analyzing and comparing the results with the baseline AI model, and drawing conclusions.
Results. Based on the analysis of dependencies, an integrated borrower default prediction methods developed and calculated. It demonstrated a significant improvement in quality metrics (an increase in average accuracy of 0.383, an increase in f1-score of 0.509, and an increase in accuracy of 0.792) relative to the baseline model. This article presents the results of experiments aimed at improving the quality metrics of machine learning models used to predict borrower default.
Conclusion. The development of integrated borrower default prediction methods will improve the accuracy and reliability of forecast models, which is of great practical importance.

Keywords: methods for reducing class imbalance, methods for isolating anomalies into a separate model, bagging method, integral method for predicting borrower default

For citation. Konstantinov A.F., Dyakonova L.P. Comparative analysis of class imbalance reduction methods in building machine learning models in the financial sector. News of the Kabardino-Balkarian Scientific Center of RAS. Vol. 27. No. 5. Pp. 68–79. DOI: 10.35330/1991-6639-2025-27-5-68-79

References

1. Information and analytical material on the development of the banking sector of the Russian Federation in December 2024. [Электронный ресурс]. Режим доступа: https://www.cbr.ru/collection/collection/file/55056/razv_bs_24_12.pdf (дата обращения: 17.09.2025). (In Russian)
2. Ali A.A., Khedr A.M., El-Bannany M., Kanakkayil S. A powerful predicting model for financial statement fraud based on optimized xgboost ensemble learning technique. Applied
   Sciences. 2023. Vol. 13. No. 4. P. 2272. DOI: 10.3390/app13042272
3. Konstantinov A.F., Dyakonova L.P. Comparative analysis of class imbalance reduction methods in building machine learning models in financial sector. News of the Kabardino-Balkarian Scientific Center of RAS. 2025. Vol. 27. No. 1. Pp. 143–151. DOI: 10.35330/1991-6639-2025-27-1-143-151. (In Russian)
4. Qian H., Zhang S., Wang B. et al. A comparative study on machine learning models combining with outlier detection and balanced sampling methods for credit scoring 2021.
   [Электронный ресурс]. Режим доступа: https://arxiv.org/abs/2112.13196 (дата обращения: 01.09.2025). DOI: 10.48550/arXiv.2112.13196
5. Dyakonova L., Konstantinov A. Approaches to risk analysis in the financial sector based on machine learning and artificial intelligence methods / MPRA Paper. [Электронный ресурс]. Режим доступа: https://mpra.ub.uni-muenchen.de/122941/ (дата обращения: 17.09.2025)
6. Liu F.T., Ting K.M., Zhou Z.H. Isolation forest. IEEE Xplore. 2008. Pp. 413–422. DOI: 10.1109/ICDM.2008.17
7. Blázquez-García A., Conde A., Mori U., Lozano J.A. A review on outlier/anomaly detection in time series data. [Электронный ресурс]. Режим доступа: https://arxiv.org/abs/2002.04236 (дата обращения: 01.09.2025).
8. Ribeiro M.T., Singh S., Guestrin C. Why should I trust you? Explaining the predictions of any classifier. [Электронный ресурс]. Режим доступа: https://arxiv.org/abs/1602.04938 (дата обращения: 01.09.2025).
9. Breiman L. Bagging predictors. Machine Learning. 1996. Vol. 24. No. 2. Pp. 123–140.
10. Abdoli M., Akbari M., Shahrabi J. Bagging supervised autoencoder classifier for credit scoring. Preprint. DOI: 10.48550/arXiv.2108.078
11. Zou Y., Gao C., Xia M., Pang C. Credit scoring based on a bagging-cascading boosted decision tree. Intelligent Data Analysis. 2022. Vol. 26. No. 6. Pp. 1557–1578. DOI: 10.3233/IDA-216228

Information about the authors

Alexey F. Konstantinov, Postgraduate Student, Department of Informatics, Plekhanov Russian University of Economics;
36 Stremyannyy lane, Moscow, 115054, Russia;
konstantinovaf@gmail.com, ORCID: https://orcid.org/0009-0000-9591-3301, SPIN-code: 3088-3121
Lyudmila P. Dyakonova, Candidate of Physical and Mathematical Sciences, Associate Professor, Department of Informatics, Plekhanov Russian University of Economics;
36 Stremyannyy lane, Moscow, 115054, Russia;
Dyakonova.LP@rea.ru, ORCID: https://orcid.org/0000-0001-5229-8070, SPIN-code: 2513-8831