A Hybrid Symbolic Regression Method for the Problem of Dynamic System Identification
Mekhatronika, Avtomatizatsiya, Upravlenie · 05.05.2026 · 27 (5) · 254-259
DOI: 10.17587/mau.24.254-259 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mekhatronika, Avtomatizatsiya, Upravlenie, online erschienen am 05.05.2026. Verfasst von L. Zhang. Die Arbeit nennt 20 Literaturangaben.
Abstract (Originalsprache)
Identifying dynamic systems from data is a complex problem, where a key requirement of modern research is not only accuracy but also model interpretability. Although highly effective, the symbolic regression method based on genetic programming has inherent limitations, the most important of which is stochasticity, leading to instability of results. In this paper, a new hybrid method, GP-SINDy, is proposed to overcome these shortcomings. Its core idea is to combine two approaches: genetic programming performs a global search for the model structure, while sparse identification fine-tunes the corresponding parameters. The effectiveness of the proposed method was validated through comprehensive computational experiments. On test data, GP-SINDy demonstrated the ability to find models with an optimal balance of accuracy and complexity, outperforming the baseline genetic programming algorithm. Analysis on noisy data confirmed the increased efficiency of the proposed method. Verification on a real system demonstrated the practical applicability of the approach for constructing adequate analytical models. Thus, the GP-SINDy hybrid method represents a powerful and versatile tool for automatically deriving interpretable system dynamics equations, opening up new possibilities in various fields of science and engineering.
Zitieren
APA
Zhang, L. (2026). A Hybrid Symbolic Regression Method for the Problem of Dynamic System Identification. Mekhatronika, Avtomatizatsiya, Upravlenie, 27(5), 254-259. https://doi.org/10.17587/mau.24.254-259
Vancouver
Zhang L. A Hybrid Symbolic Regression Method for the Problem of Dynamic System Identification. Mekhatronika, Avtomatizatsiya, Upravlenie. 2026;27(5):254-259. doi:10.17587/mau.24.254-259
BibTeX
@article{Zhang2026,
author = {Zhang, L.},
title = {A Hybrid Symbolic Regression Method for the Problem of Dynamic System Identification},
journal = {Mekhatronika, Avtomatizatsiya, Upravlenie},
year = {2026},
volume = {27},
number = {5},
pages = {254--259},
doi = {10.17587/mau.24.254-259}
}
EUVIMED