Experimental evaluation of a self-learning recommender system for the banking sector
DOI: 10.21293/1818-0442-2026-29-1-160-165
Abstract: Relevance. In the banking sector, personalized recommendations must remain accurate, robust, and safe under changing customer behavior and financial risks. It is particularly important to test such systems under macroeconomic stress, when conventional models may lose performance. Purpose. The study evaluates a prototype self-learning recommender system for banking products that combines reinforcement learning, macroeconomic context, and anomaly filtering. Methods. The work uses experimental simulation, comparative analysis with baseline algorithms, k-means customer clustering, deep Q-learning, isolation forest anomaly detection, and statistical assessment of P@5, NDCG@5, conversion, F1-score, and recommendation latency. Novelty. The study proposes an integrated approach, in which an adaptive recommendation policy is combined with macroeconomic regimes awareness and preliminary anomaly detection for banking use cases. Results. In the synthetic benchmark, the prototype outperformed baseline models: P@5 reached 0,623, NDCG@5 reached 0,712, conversion in the stable regime was 82%, and the performance drop under crisis conditions was limited to 7%. The anomaly detection module achieved an F1-score of at least 0,86, while the average recommendation latency was 23 ms. Practical significance. The results can support the design of banking personalization systems, selection of recommender architectures, robustness assessment, and implementation of risk-control mechanisms in digital customer service channels.
Keywords: recommender systems, deep reinforcement learning, banking sector, personalization, macroeconomic scenarios, anomaly detection, quality metrics
For citation:
Askerov Z. H. Experimental evaluation of a self-learning recommender system for the banking sector. Doklady Tomskogo gosudarstvennogo universiteta sistem upravleniya i radioelektroniki, 2026, vol. 29, no. 1, pp. 160–165. DOI: 10.21293/1818-0442-2026-29-1-160-165
Authors and copyright holders:
- Askerov Z. H. , National Research Nuclear University MEPhI, (Moscow Engineering Physics Institute), 31, Kashirskoe Hwy, Moscow
- 1. Gigli A., Lillo F., Regoli D. Recommender systems for banking and financial services. Poster Track of the 11th ACM Conference on Recommender Systems (RecSys 2017). Como, Italy, CEUR-WS.org, 2017, pp. 315–316.
- 2. Oyebode O., Orji R. A hybrid recommendation system for the financial service sector. Journal of Banking and Financial Technology, 2020, vol. 4, pp. 17–32.
- 3. Sharaf M., Hemdan E.E.D., El-Sayed A., El-Bahnasawy N.A. A survey on recommendation systems for financial services. Multimedia Tools and Applications, 2022, vol. 81, no. 12, pp. 13617–13642.
- 4. Vaquero-Patricio C., van Ommeren N., Gil-Begue S. Recommenders in Banking: An End-to-End Personalization Pipeline within ING. Proceedings of the 15th ACM Conference on Recommender Systems (RecSys 2021). Amsterdam, Netherlands, ACM, 2021, pp. 587–589. DOI 10.1145/3460231.3474612.
- 5. Afsar M.M., Crump T., Far B. Reinforcement learning based recommender systems: A survey. ACM Computing Surveys, 2021, vol. 54, no. 3, article 42, pp. 1–38.
- 6. Chen M., Beutel A., Covington P., Jain S., Belletti F., Chi E. Top-K off-policy correction for a REINFORCE recommender system. Proceedings of the 12th ACM International Conference on Web Search and Data Mining (WSDM 2019). Melbourne, VIC, Australia, ACM, 2019, pp. 456–464. DOI: 10.1145/3289600.3290999.
- 7. Mnih V., Kavukcuoglu K., Silver D. et al. Human-level control through deep reinforcement learning. Nature, 2015, vol. 518, no. 7540, pp. 529–533.
- 8. Wang Z., Schaul T., Hessel M. et al. Dueling Network Architectures for Deep Reinforcement Learning. Proceedings of the 33rd International Conference on Machine Learning (ICML 2016). New York, NY, USA, PMLR, 2016, pp. 1995–2003.
- 9. Kutanin A.R. [Recommender systems: survey of main problem statements and results]. Intelligent systems. Theory and Applications, 2017, vol. 21, no. 4, pp. 18–30 (in Russ.).
- 10. Menshikova N.V., Portnov I.V., Nikolaev I.E. [Review of recommender systems and opportunities for context accounting in individual recommendations]. Academy, 2016, no. 6 (9), pp. 20–22 (in Russ.).
- 11. Pavlov D.A., Kosyuga O.S. [A universal recommender system for banking products based on a hybrid approach]. Scientific Papers of Kuban State Technological University, 2022, no. 1, pp. 88–97 (in Russ.).
- 12. Arkadyeva O.G., Petrov A.V. [Machine learning models as a tool for detecting suspicious bank operations]. Vestnik Surgutskogo gosudarstvennogo universiteta, 2025, vol. 13, no. 3, pp. 8-21. DOI: 10.35266/2949-3455-2025-3-1 (in Russ.).
- 13. Koren Y., Bell R., Volinsky C. Matrix factorization techniques for recommender systems. IEEE Computer, 2009, vol. 42, no. 8, pp. 30–37.
- 14. Rendle S., Freudenthaler C., Gantner Z., Schmidt-Thieme L. BPR: Bayesian Personalized Ranking from Implicit Feedback. Proceedings of the 25th Conference on Uncertainty in Artificial Intelligence (UAI 2009). Montreal, QC, Canada, AUAI Press, 2009, pp. 452–461.
- 15. Järvelin K., Kekäläinen J. Cumulated gain-based evaluation of IR techniques. ACM Transactions on Information Systems, 2002, vol. 20, no. 4, pp. 422–446.
- 16. Liu F.T., Ting K.M., Zhou Z.-H. Isolation Forest. Proceedings of the 8th IEEE International Conference on Data Mining (ICDM 2008). Pisa, Italy, IEEE, 2008, pp. 413–422. DOI: 10.1109/ICDM.2008.17.
- 17. Hariri S., Carrasco Kind M., Brunner R.J. Extended Isolation Forest. IEEE Transactions on Knowledge and Data Engineering, 2021, vol. 33, no. 4, pp. 1479–1489.
- 18. Pang G., Shen C., Cao L., van den Hengel A. Deep learning for anomaly detection: A review. ACM Computing Surveys, 2021, vol. 54, no. 2, article 38, pp. 1–38.
- 19. Hilal W., Gadsden S.A., Yawney J. Financial fraud: A review of anomaly detection techniques and recent advances. Expert Systems with Applications, 2022, vol. 193, article 116429.
- 20. Bifet A., Gavaldà R. Learning from time-changing data with adaptive windowing. Proceedings of the 7th SIAM International Conference on Data Mining (SDM 2007). Minneapolis, MN, USA, SIAM, 2007, pp. 443–448. DOI: 10.1137/1.9781611972771.42.
- 21. Lu J., Liu A., Dong F., Gama J., Zhang G. Learning under concept drift: A review. IEEE Transactions on Knowledge and Data Engineering, 2019, vol. 31, no. 12, pp. 2346–2363.
- 22. Hoi S.C.H., Sahoo D., Lu J., Zhao P. Online learning: A comprehensive survey. Neurocomputing, 2021, vol. 459, pp. 249–289.
- 23. Raab C., Ziebart B., Kellermeier M., Schelter S. Reactive and self-adjusting sliding windows for efficient event stream processing with general delay distributions. Neurocomputing, 2020, vol. 416, pp. 340–351.
- 24. De Lange M., Aljundi R., Masana M. et al. A continual learning survey: Defying forgetting in classification tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, vol. 44, no. 7, pp. 3366–3385.
- 25. Bank of England; Financial Conduct Authority. Machine learning in UK financial services. Report, October 2022. Available at: https://www.bankofengland.co.uk/report/2022/machine-learning-in-uk-financial-services (accessed: 26 February 2026).
- 26. Nystrup P., Lindström E., Christiansen T., Madsen H., Engsø A. Detecting change points in finance: A mid-2020 review. Expert Systems with Applications, 2022, vol. 77, no. 1, pp. 5–47.
- 27. Fortunato M., Azar M.G., Piot B. et al. Noisy Networks for Exploration. Proceedings of the International Conference on Learning Representations (ICLR 2018). Vancouver, BC, Canada, OpenReview.net, 2018, 13 p. Available at: https://openreview.net/forum?id=rywHCPkAW (accessed: 26 February 2026).
Last issue
-2026-1_%D0%BA%D0%BE%D0%BF%D0%B8%D1%8F.jpg)
Editorial office address
Executive Secretary of the Editor’s Office
Editor’s Office: 40 Lenina Prospect, Tomsk, 634050, Russia
Phone / Fax: + 7 (3822) 701-582
Viktor N. Maslennikov
Executive Secretary of the Editor’s Office
Editor’s Office: 40 Lenina Prospect, Tomsk, 634050, Russia
Phone / Fax: + 7 (3822) 51-21-21 / 51-43-02
Subscription for updates
