EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

From data-driven differential equations to mechanistic forecasting with machine learning

Amit K Chakraborty iD, Hao Wang iD, Pouria Ramazi iD

Journal of the Royal Society Interface · 04.11.2026 · 23 (244)

DOI: 10.1098/rsif.2026.0240 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of the Royal Society Interface, online erschienen am 04.11.2026. Verfasst von Amit K Chakraborty und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Mathematical and Statistical Sciences, University of Alberta , Edmonton, Alberta, Canada. Hauptthemen laut MEDLINE: Machine Learning. Die Arbeit nennt 89 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Natural Sciences and Engineering Research Council of Canada, Canada Research Chairs programme, NSERC Discovery Grant.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The equations of complex dynamical systems may not be identified by expert knowledge, especially if the underlying mechanisms are unknown. Data-driven discovery methods address this challenge by inferring governing equations from time-series data using a library of functions constructed from the measured variables. However, these methods typically assume time-invariant coefficients, which limits their ability to capture evolving system dynamics. To overcome this limitation, we allow some of the parameters to vary over time, learn their temporal evolution directly from data and infer a system of equations that incorporates both constant and time-varying parameters. We then transform this framework into a forecasting model by predicting the time-varying parameters and substituting these predictions into the learned equations. The model is validated on simulated susceptible–infected–recovered and consumer–resource systems, as well as on real-world greenhouse gas concentration and cyanobacteria cell count datasets. By dynamically adapting to temporal shifts, our proposed model achieved a mean absolute error below 3% for learning a time series and below 6% for forecasting up to a month ahead. We additionally compare forecasting performance against convolutional neural network–long short-term memory and gradient boosting machine, and show that our model outperforms these methods across most datasets. Our findings demonstrate that integrating time-varying parameters into data-driven discovery of differential equations improves both modelling accuracy and forecasting performance.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Machine LearningForecastingPredictive Learning ModelsPrediction AlgorithmsBoosting Machine Learning Algorithms

Stichwörter der Autor:innen

data-driven dynamical systemsfinite-horizon forecast error boundsmachine learningmechanistic forecastingsparse regressiontime-varying parameters

Autor:innen und Institutionen

  • Department of Mathematical and Statistical Sciences, University of Alberta , Edmonton, Alberta, Canada.
  • Department of Biological Sciences and Department of Mathematics and Statistics, University of Calgary , Calgary, Alberta, Canada.

ORCID-Profile: Amit K Chakraborty, Hao Wang, Pouria Ramazi

Förderung

  • Natural Sciences and Engineering Research Council of Canada
  • Canada Research Chairs programme
  • NSERC Discovery Grant

Zitieren

APA

Chakraborty, A. K., Wang, H., & Ramazi, P. (2026). From data-driven differential equations to mechanistic forecasting with machine learning. Journal of the Royal Society Interface, 23(244). https://doi.org/10.1098/rsif.2026.0240

Vancouver

Chakraborty AK, Wang H, Ramazi P. From data-driven differential equations to mechanistic forecasting with machine learning. Journal of the Royal Society Interface. 2026;23(244). doi:10.1098/rsif.2026.0240

BibTeX
@article{Chakraborty2026,
  author = {Chakraborty, Amit K and Wang, Hao and Ramazi, Pouria},
  title = {From data-driven differential equations to mechanistic forecasting with machine learning},
  journal = {Journal of the Royal Society Interface},
  year = {2026},
  volume = {23},
  number = {244},
  doi = {10.1098/rsif.2026.0240}
}
Kostenlos starten →