Риск-ориентированная маршрутизация моделей при обработке банковских транзакций
Аннотация
Обоснование. В банковском антифроде рост точности моделей нередко сопровождается увеличением времени инференса, стоимости вычислений и нагрузки на ручную проверку. Единый вычислительный маршрут для всех операций способен ухудшать хвостовую задержку и повышать операционный риск при жестком сроке авторизации.
Цель. Разработать и оценить риск-ориентированную архитектуру выбора модели искусственного интеллекта, которая согласует глубину проверки с предварительным риском операции, достоверностью признаков и остаточным бюджетом времени.
Материалы и методы. Транзакционное решение формализовано как выбор маршрута при асимметричной цене пропуска мошенничества, ложного отказа, нарушения соглашения об уровне обслуживания, вычислений и ручного разбора. Проведен имитационный эксперимент: 30 независимых запусков по 100 000 операций, базовый и стрессовый сценарии, сравнение монолитного конвейера, прямой легкой модели и селективной маршрутизации. Оценивались средняя задержка, 95-й и 99-й процентили, доля ответов свыше 200 мс и доверительные интервалы уровня 95%; выполнены анализ чувствительности и абляции.
Результаты. В стрессовом сценарии средний 99-й процентиль уменьшился с 254,9 до 99,1 мс, а доля превышений 200 мс - с 25,422 до 0,133%. При доле усиленного маршрута 5% целевой порог сохранялся, при 10% p99 достигал 205,2 мс. Результаты показывают резерв операционной эффективности, но не оценивают полноту выявления мошенничества и денежный эффект; эти показатели требуют проверки на размеченных транзакциях и производственных трассах.
Скачивания
Литература
Недбаев, А. А., & Спиридонов, С. П. (2025). Обеспечение безопасности в банковских транзакциях при помощи искусственного интеллекта. Путеводитель предпринимателя, 18(3), 20-25. https://doi.org/10.24182/2073-9885-2025-18-3-20-25. EDN: https://elibrary.ru/STBQOP
Akidau, T., Bradshaw, R., Chambers, C., Chernyak, S., Fernández-Moctezuma, R. J., Lax, R., McVeety, S., Mills, D., Perry, F., Schmidt, E., & Whittle, S. (2015). The Dataflow Model: A practical approach to balancing correctness, latency, and cost in massive-scale, unbounded, out-of-order data processing. Proceedings of the VLDB Endowment, 8(12), 1792-1803. https://doi.org/10.14778/2824032.2824076
Baesens, B., Höppner, S., & Verdonck, T. (2021). Data engineering for fraud detection. Decision Support Systems, 150, Article 113492. https://doi.org/10.1016/j.dss.2021.113492. EDN: https://elibrary.ru/PYZLLU
Carcillo, F., Dal Pozzolo, A., Le Borgne, Y.-A., Caelen, O., Mazzer, Y., & Bontempi, G. (2018). SCARFF: A scalable framework for streaming credit card fraud detection with Spark. Information Fusion, 41, 182-194. https://doi.org/10.1016/j.inffus.2017.09.005
Carcillo, F., Le Borgne, Y.-A., Caelen, O., Kessaci, Y., Oblé, F., & Bontempi, G. (2021). Combining unsupervised and supervised learning in credit card fraud detection. Information Sciences, 557, 317-331. https://doi.org/10.1016/j.ins.2019.05.042. EDN: https://elibrary.ru/NPYVSQ
Cheng, T., & Xiao, J. (2025). Cost-constrained feature selection using gradient boosting decision tree. Engineering Applications of Artificial Intelligence, 162, Article 112751. https://doi.org/10.1016/j.engappai.2025.112751. EDN: https://elibrary.ru/ZOSHYN
Crankshaw, D., Wang, X., Zhou, G., Franklin, M. J., Gonzalez, J. E., & Stoica, I. (2017). Clipper: A low-latency online prediction serving system. In Proceedings of the 14th USENIX Symposium on Networked Systems Design and Implementation (pp. 613-627). USENIX Association. https://www.usenix.org/conference/nsdi17/technical-sessions/presentation/crankshaw.
Crankshaw, D., Sela, G.-E., Mo, X., Zumar, C., Stoica, I., Gonzalez, J. E., & Tumanov, A. (2020). InferLine: Latency-aware provisioning and scaling for prediction serving pipelines. In Proceedings of the 11th ACM Symposium on Cloud Computing (pp. 477-491). Association for Computing Machinery. https://doi.org/10.1145/3419111.3421285
Dal Pozzolo, A., Boracchi, G., Caelen, O., Alippi, C., & Bontempi, G. (2015). Credit card fraud detection and concept-drift adaptation with delayed supervised information. In 2015 International Joint Conference on Neural Networks (pp. 1-8). Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/IJCNN.2015.7280527
Dal Pozzolo, A., Boracchi, G., Caelen, O., Alippi, C., & Bontempi, G. (2018). Credit card fraud detection: A realistic modeling and a novel learning strategy. IEEE Transactions on Neural Networks and Learning Systems, 29(8), 3784-3797. https://doi.org/10.1109/TNNLS.2017.2736643
Dean, J., & Barroso, L. A. (2013). The tail at scale. Communications of the ACM, 56(2), 74-80. https://doi.org/10.1145/2408776.2408794
Gujarati, A., Karimi, R., Alzayat, S., Hao, W., Kaufmann, A., Vigfusson, Y., & Mace, J. (2020). Serving DNNs like Clockwork: Performance predictability from the bottom up. In Proceedings of the 14th USENIX Symposium on Operating Systems Design and Implementation (pp. 443-462). USENIX Association. https://www.usenix.org/conference/osdi20/presentation/gujarati.
Höppner, S., Baesens, B., Verbeke, W., & Verdonck, T. (2022). Instance-dependent cost-sensitive learning for detecting transfer fraud. European Journal of Operational Research, 297(1), 291-300. https://doi.org/10.1016/j.ejor.2021.05.028. EDN: https://elibrary.ru/PUBSDR
Huang, D., Drietomsky, T., Barrett, B., & Wang, Z. (2026). How small can you go? LoRA fine-tuning 270M-8B models for merchant information extraction in financial transactions. arXiv. https://doi.org/10.48550/arXiv.2606.08051
Olston, C., Fiedel, N., Gorovoy, K., Harmsen, J., Lao, L., Li, F., Rajashekhar, V., Ramesh, S., & Soyke, J. (2017). TensorFlow-Serving: Flexible, high-performance ML serving. arXiv. https://doi.org/10.48550/arXiv.1712.06139
Plakandaras, V., Gogas, P., Papadimitriou, T., & Tsamardinos, I. (2022). Credit card fraud detection with automated machine learning systems. Applied Artificial Intelligence, 36(1), Article 2086354. https://doi.org/10.1080/08839514.2022.2086354. EDN: https://elibrary.ru/QASOON
Prasad, A., Mofjeld, C., & Peng, Y. (2021). A joint model provisioning and request dispatch solution for low-latency inference services on edge. Sensors, 21(19), Article 6594. https://doi.org/10.3390/s21196594. EDN: https://elibrary.ru/CXBHAS
Romero, F., Li, Q., Yadwadkar, N. J., & Kozyrakis, C. (2021). INFaaS: Automated model-less inference serving. In Proceedings of the 2021 USENIX Annual Technical Conference (pp. 397-411). USENIX Association. https://www.usenix.org/conference/atc21/presentation/romero.
Sculley, D., Holt, G., Golovin, D., Davydov, E., Phillips, T., Ebner, D., Chaudhary, V., Young, M., Crespo, J.-F., & Dennison, D. (2015). Hidden technical debt in machine learning systems. In Advances in Neural Information Processing Systems 28 (pp. 2503-2511). Curran Associates. https://proceedings.neurips.cc/paper_files/paper/2015/hash/86df7dcfd896fcaf2674f757a2463eba-Abstract.html.
Sreekanti, V., Subbaraj, H., Wu, C., Gonzalez, J. E., & Hellerstein, J. M. (2020). Optimizing prediction serving on low-latency serverless dataflow. arXiv. https://doi.org/10.48550/arXiv.2007.05832
Tabassi, E. (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0) (NIST AI 100-1). National Institute of Standards and Technology. https://doi.org/10.6028/NIST.AI.100-1
Vanderschueren, T., Baesens, B., Verdonck, T., & Verbeke, W. (2024). A new perspective on classification: Optimally allocating limited resources to uncertain tasks. Decision Support Systems, 179, Article 114151. https://doi.org/10.1016/j.dss.2023.114151. EDN: https://elibrary.ru/AEJTCW
Wan, C., Santriaji, M., Rogers, E., Hoffmann, H., Maire, M., & Lu, S. (2020). ALERT: Accurate learning for energy and timeliness. In Proceedings of the 2020 USENIX Annual Technical Conference (pp. 353-369). USENIX Association. https://www.usenix.org/conference/atc20/presentation/wan.
Wang, D., Lin, J., Cui, P., Jia, Q., Wang, Z., Fang, Y., Yu, Q., Zhou, J., Yang, S., & Qi, Y. (2019). A semi-supervised graph attentive network for financial fraud detection. In 2019 IEEE International Conference on Data Mining (pp. 598-607). Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/ICDM.2019.00070
References
Nedbaev, A. A., & Spiridonov, S. P. (2025). Ensuring security in banking transactions using artificial intelligence. Entrepreneur's Guide, 18(3), 20–25. https://doi.org/10.24182/2073-9885-2025-18-3-20-25. EDN: https://elibrary.ru/STBQOP
Akidau, T., Bradshaw, R., Chambers, C., Chernyak, S., Fernández-Moctezuma, R. J., Lax, R., McVeety, S., Mills, D., Perry, F., Schmidt, E., & Whittle, S. (2015). The Dataflow Model: A practical approach to balancing correctness, latency, and cost in massive-scale, unbounded, out-of-order data processing. Proceedings of the VLDB Endowment, 8(12), 1792–1803. https://doi.org/10.14778/2824032.2824076
Baesens, B., Höppner, S., & Verdonck, T. (2021). Data engineering for fraud detection. Decision Support Systems, 150, Article 113492. https://doi.org/10.1016/j.dss.2021.113492. EDN: https://elibrary.ru/PYZLLU
Carcillo, F., Dal Pozzolo, A., Le Borgne, Y.-A., Caelen, O., Mazzer, Y., & Bontempi, G. (2018). SCARFF: A scalable framework for streaming credit card fraud detection with Spark. Information Fusion, 41, 182–194. https://doi.org/10.1016/j.inffus.2017.09.005
Carcillo, F., Le Borgne, Y.-A., Caelen, O., Kessaci, Y., Oblé, F., & Bontempi, G. (2021). Combining unsupervised and supervised learning in credit card fraud detection. Information Sciences, 557, 317–331. https://doi.org/10.1016/j.ins.2019.05.042. EDN: https://elibrary.ru/NPYVSQ
Cheng, T., & Xiao, J. (2025). Cost-constrained feature selection using gradient boosting decision tree. Engineering Applications of Artificial Intelligence, 162, Article 112751. https://doi.org/10.1016/j.engappai.2025.112751. EDN: https://elibrary.ru/ZOSHYN
Crankshaw, D., Wang, X., Zhou, G., Franklin, M. J., Gonzalez, J. E., & Stoica, I. (2017). Clipper: A low-latency online prediction serving system. In Proceedings of the 14th USENIX Symposium on Networked Systems Design and Implementation (pp. 613–627). USENIX Association. https://www.usenix.org/conference/nsdi17/technical-sessions/presentation/crankshaw
Crankshaw, D., Sela, G.-E., Mo, X., Zumar, C., Stoica, I., Gonzalez, J. E., & Tumanov, A. (2020). InferLine: Latency-aware provisioning and scaling for prediction serving pipelines. In Proceedings of the 11th ACM Symposium on Cloud Computing (pp. 477–491). Association for Computing Machinery. https://doi.org/10.1145/3419111.3421285
Dal Pozzolo, A., Boracchi, G., Caelen, O., Alippi, C., & Bontempi, G. (2015). Credit card fraud detection and concept-drift adaptation with delayed supervised information. In 2015 International Joint Conference on Neural Networks (pp. 1–8). Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/IJCNN.2015.7280527
Dal Pozzolo, A., Boracchi, G., Caelen, O., Alippi, C., & Bontempi, G. (2018). Credit card fraud detection: A realistic modeling and a novel learning strategy. IEEE Transactions on Neural Networks and Learning Systems, 29(8), 3784–3797. https://doi.org/10.1109/TNNLS.2017.2736643
Dean, J., & Barroso, L. A. (2013). The tail at scale. Communications of the ACM, 56(2), 74–80. https://doi.org/10.1145/2408776.2408794
Gujarati, A., Karimi, R., Alzayat, S., Hao, W., Kaufmann, A., Vigfusson, Y., & Mace, J. (2020). Serving DNNs like Clockwork: Performance predictability from the bottom up. In Proceedings of the 14th USENIX Symposium on Operating Systems Design and Implementation (pp. 443–462). USENIX Association. https://www.usenix.org/conference/osdi20/presentation/gujarati
Höppner, S., Baesens, B., Verbeke, W., & Verdonck, T. (2022). Instance-dependent cost-sensitive learning for detecting transfer fraud. European Journal of Operational Research, 297(1), 291–300. https://doi.org/10.1016/j.ejor.2021.05.028. EDN: https://elibrary.ru/PUBSDR
Huang, D., Drietomsky, T., Barrett, B., & Wang, Z. (2026). How small can you go? LoRA fine-tuning 270M-8B models for merchant information extraction in financial transactions. arXiv. https://doi.org/10.48550/arXiv.2606.08051
Olston, C., Fiedel, N., Gorovoy, K., Harmsen, J., Lao, L., Li, F., Rajashekhar, V., Ramesh, S., & Soyke, J. (2017). TensorFlow-Serving: Flexible, high-performance ML serving. arXiv. https://doi.org/10.48550/arXiv.1712.06139
Plakandaras, V., Gogas, P., Papadimitriou, T., & Tsamardinos, I. (2022). Credit card fraud detection with automated machine learning systems. Applied Artificial Intelligence, 36(1), Article 2086354. https://doi.org/10.1080/08839514.2022.2086354. EDN: https://elibrary.ru/QASOON
Prasad, A., Mofjeld, C., & Peng, Y. (2021). A joint model provisioning and request dispatch solution for low-latency inference services on edge. Sensors, 21(19), Article 6594. https://doi.org/10.3390/s21196594. EDN: https://elibrary.ru/CXBHAS
Romero, F., Li, Q., Yadwadkar, N. J., & Kozyrakis, C. (2021). INFaaS: Automated model-less inference serving. In Proceedings of the 2021 USENIX Annual Technical Conference (pp. 397–411). USENIX Association. https://www.usenix.org/conference/atc21/presentation/romero
Sculley, D., Holt, G., Golovin, D., Davydov, E., Phillips, T., Ebner, D., Chaudhary, V., Young, M., Crespo, J.-F., & Dennison, D. (2015). Hidden technical debt in machine learning systems. In Advances in Neural Information Processing Systems 28 (pp. 2503–2511). Curran Associates. https://proceedings.neurips.cc/paper_files/paper/2015/hash/86df7dcfd896fcaf2674f757a2463eba-Abstract.html
Sreekanti, V., Subbaraj, H., Wu, C., Gonzalez, J. E., & Hellerstein, J. M. (2020). Optimizing prediction serving on low-latency serverless dataflow. arXiv. https://doi.org/10.48550/arXiv.2007.05832
Tabassi, E. (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0) (NIST AI 100-1). National Institute of Standards and Technology. https://doi.org/10.6028/NIST.AI.100-1
Vanderschueren, T., Baesens, B., Verdonck, T., & Verbeke, W. (2024). A new perspective on classification: Optimally allocating limited resources to uncertain tasks. Decision Support Systems, 179, Article 114151. https://doi.org/10.1016/j.dss.2023.114151. EDN: https://elibrary.ru/AEJTCW
Wan, C., Santriaji, M., Rogers, E., Hoffmann, H., Maire, M., & Lu, S. (2020). ALERT: Accurate learning for energy and timeliness. In Proceedings of the 2020 USENIX Annual Technical Conference (pp. 353–369). USENIX Association. https://www.usenix.org/conference/atc20/presentation/wan
Wang, D., Lin, J., Cui, P., Jia, Q., Wang, Z., Fang, Y., Yu, Q., Zhou, J., Yang, S., & Qi, Y. (2019). A semi-supervised graph attentive network for financial fraud detection. In 2019 IEEE International Conference on Data Mining (pp. 598–607). Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/ICDM.2019.00070
Copyright (c) 2026 Georgii V. Bolshakov, Pavel I. Kozlov, Rimma Ya. Labkovskaya
Это произведение доступно по лицензии Creative Commons «Attribution-NonCommercial-NoDerivatives» («Атрибуция — Некоммерческое использование — Без производных произведений») 4.0 Всемирная.
