L.S. Zvyagin1
1 Financial University under the Government of the Russian Federation (Moscow, Russia)
1 lszvyagin@fa.ru

Modern information systems (IS) face the problem of instability and low adaptability when operating in conditions of noisy data, high uncertainty and massive cyber attacks (DDoS). Traditional deterministic algorithms and classical filtering methods often turn out to be ineffective, allowing degradation of services and "information collapse" at peak loads. The article solves the actual problem of inefficiency of traditional algorithms for managing information systems (IS) in conditions of noisy data and high uncertainty.

Development and substantiation of an IP management methodology based on a regularizing Bayesian approach that provides a quan titative assessment of uncertainty and adaptability of the system.

The developed model, unlike existing approaches, provides quantitative transfer functions between levels in order to evaluate in real time how technical parameters (for example, server CPU usage or database response time) affect business process metrics (for example, order completion time) and, ultimately, on strategic KPIs (for example, conversion or LTV of the client). The article provides a detailed mathematical model for calculating the integrated efficiency of an IP, and demonstrates its application using a computational example for an e-commerce system.

Mathematical modeling of IP dynamics has been carried out using systems of nonlinear differential equations (predator-prey models) describing the interaction of legitimate and legitimate traffic. To verify the proposed approach, a numerical experiment was carried out using the Runge-Kutta method of the 4th order, in which the scenarios of the system without control, with the classical Kalman-Busey filter and with the author's regularizing Bayesian algorithm were compared. The results of the study demonstrated the supe­riority of the proposed methodology. The scientific novelty of the work lies in the integration of machine learning methods with the theory of differential games and the incorporation of the regularizing function into the structure of the Bayesian approach.

Zvyagin L.S. Management of information systems in conditions of high uncertainty using a Bayesian approach // Neurocomputers. 2026. V. 28. № 3. P. 15–24. DOI: https://doi.org/10.18127/j19998554-202603-03.

1. Akram M., Adnan M., Ali S.F., Ahmad J., Youssef A., Alshalali T.A.N., Sheikh Z.A. Detection of diabetic retinopathy taking into account uncertainty using deep learning enhanced by Bayesian approaches. Scientific Reports. 2025. V. 15. № 1. P. 1342.
2. Garcia S., Mora-Merchant H.M., Larios D.F., Staff E., Parejo A., Leon S. Identification of phase topology in low-voltage distribution networks: a Bayesian approach. International Journal of Electric Power Engineering and Energy Systems. 2023. V. 144. P. 108525.
3. Hong K.K., Song S., Xia B., Jimmy N.L., Wei K.A., Wu P.P.Y. Application of Bayesian approaches in construction management research: a systematic review. Engineering, construction and architectural management. 2022. V. 29. № 5. P. 2153–2182.
4. Kiani A.K., Khan V.U., Raja M.A.Z., On Yu., Sabir Z., Shoaib M. Intelligent backpropagation networks with Bayesian regularization for mathematical models of ecological and economic systems. Sustainable Development. 2021. V. 13. № 17. P. 9537.
5. Liu R.L., Hua Yu., Zhou Z.F., Li Yu., Wu U.T., Aubrey N. Forecasting and optimization of airfoil aerodynamic characteristics using the Bayesian method associated with a deep neural network. Physics of fluids. 2022. V. 34. № 11. P. 114–114.
6. Mohammad-Jafari A. Regularization, Bayesian inference and machine learning methods for inverse problems. Entropy. 2021. V. 23.
   № 12. P. 1673.
7. Ndife A.N., Mensin Yu., Rakvichyan V., Munisavang P. Cybersecurity audit of Smart Grid networks: an optimized detection method based on Bayesian deep learning. Journal of Internet Services and Information Security. 2022. V. 12. № 2. P. 95–114.
8. Setiaji P., Vijanarko B., Syakhedin Y., Himavan H., Siregar N.A., Harmayani. Kartikaningsih R. The best model and effectiveness of the Bayesian regularization method for data forecasting. Physical Journal: Conference Series. 2021. V. 1933. № 1. P. 012030.
9. Song K., Sun Y., Han G., Rodriguez J. Intrusion detection based on hybrid classifiers for intelligent networks. Computers and Electrical Engineering. 2021. V. 93. P. 107212.
10. Sujata R., Mareswari V., Chatterjee J.M., Musa A.A., Hassanien A.E. Bayesian regularized neural network for analyzing Bitcoin trends. IEEE Access. 2021. V. 9. P. 37989–38000.
11. Teng D., Bhoyar K.K. Feature selection methods for machine learning: a review of more than two decades of research. Knowledge and Information Systems. 2024. V. 66. № 3. P. 1575–1637.
12. Van Erp S. Bayesian regularized SEM: modern possibilities and limitations. Psychology. 2023. V. 5. № 3. P. 814–835.
13. Veluru S.P. Self-tuning neural networks: embedded regularization using internal feedback on trust. International Journal of New Trends in Computer Science and Information Technology. 2023. V. 4. № 3. P. 41–49.
14. Wang Z.K., Hua Yu., Aubrey N., Zhou Z.F., Feng F., Wu U.T. Fast optimization of multi-chip modules using deep learning in combination with the Bayesian method. International communications in the field of heat and mass transfer. 2023. V. 141. P. 106592.
15. Zhang H., Liu H., Zhou S., Ma N. Risk management related to big data in medicine: a systematic approach to management based on Bayesian belief networks. Mobile Information Systems. 2023. V. 2023. № 1. P. 9507349.