Forecasting the load on a microservice system using the ARIMA method and Bayesian networks

The article presents an approach to predicting the load on a microservice system that combines ARIMA time series analysis methods and probabilistic inference in Bayesian networks. This approach allows for the consideration of both the load patterns on individual microservices over time and the structural dependencies between these microservices. The presented approach consists of two stages: in the first stage, ARIMA models build independent forecasts for each microservice, and in the second stage, a Bayesian network adjusts the obtained forecasts, taking into account dependencies between microservices and the propagation of load from service to service. The final forecast consists of the weighted results of both stages. In addition, an anomaly detection criterion is provided, which allows the forecasting system to respond to anomalies by changing the weights and other parameters of the algorithm. The approach is experimentally tested using real-world system data. The results are compared with the isolated use of ARIMA and the use of Long-Short Term Memory (LSTM) networks for the same task. The algorithm shows promise for use in load forecasting.

1. Grunichev Yu.A., Konyashkin R.A. Cloud computing and its impact on the architecture of modern information systems. Paradigm. 2026;(1-1):85–89. (In Russ.).

2. Box G.E.P., Jenkins G.M., Reinsel G.C., Ljung G.M. Time Series Analysis: Forecasting and Control. Hoboken: John Wiley & Sons; 2015. 720 p.

3. Saharov D.V., Gelfand A.M., Kazantsev A.A., Pestov I.E. Using mathematical forecasting methods to estimate the load on the computing power of the IoT network. Vestnik Saint-Petersburg University of State Fire Service of Emercom of Russia. 2020;(2):86–94. (In Russ.).

4. Cleveland R.B., Cleveland W.S., McRae J.E., Terpenning I. STL: A seasonal-trend decomposition procedure based on loess. Journal of Official Statistics. 1990;6(1):3–73.

5. Hochreiter S., Schmidhuber J. Long Short-Term Memory. Neural Computation. 1997;9(8):1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735

6. Breiman L. Random forests. Machine Learning. 2001;45(1):5–32. https://doi.org/10.1023/A:1010933404324

7. Friedman J.H. Greedy function approximation: A gradient boosting machine. Annals of Statistics. 2001;29(5):1189–1232. https://doi.org/10.1214/aos/1013203451

8. Terekhin M.A., Ivaschenko A.V., Kulakov G.A. A conceptual approach to the integration of artificial intelligence into engineering activities. Modeling, Optimization and Information Technology. 2025;13(2). (In Russ.). https://doi.org/10.26102/2310-6018/2025.49.2.031

9. Rabiner L.R. A tutorial on hidden Markov models and selected applications in speech recognition. Proceedings of the IEEE. 1989;77(2):257–286. https://doi.org/10.1109/5.18626

10. Koller D., Friedman N. Probabilistic Graphical Models: Principles and Techniques. Cambridge: MIT Press; 2009. 1270 p.