Data-Driven Discovery of Mathematical Models Using Artificial Intelligence
Computer and Decision Making: An International Journal · 31.10.2026 · 3 (6S) · 949-956
DOI: 10.1922/comdem.v3i6s.18769 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 31.10.2026. Verfasst von Sanjay Kumar und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The discovery of governing mathematical equations directly from observational data represents a pivotal challenge in modern computational science. Traditional model derivation relies on prior physical knowledge and manual hypothesis testing, processes that are increasingly impractical for complex, high-dimensional systems. This paper presents an AI-Driven Mathematical Discovery (AI-DM) framework that integrates deep learning, sparse regression, Gaussian process inference, and symbolic regression to autonomously identify governing differential equations from noisy, limited observational data. The proposed framework employs a four-stage pipeline: feature engineering with physics-informed normalization, latent-space representation learning via variational autoencoders, sparse identification of nonlinear dynamics (SINDy), and symbolic refinement through genetic programming. Evaluated on five canonical partial differential equations the Lorenz system, heat equation, Burgers' equation, Korteweg–de Vries (KdV) equation, and Navier-Stokes equations AI-DM achieves a mean R² of 0.944 and RMSE of 0.039, outperforming state-of-the-art baselines including Physics-Informed Neural Networks (PINNs) and classical SINDy. Discovered equations exhibit coefficient-level accuracy within 0.87% relative error. The framework operates without prior knowledge of equation structure and remains robust under 15% additive Gaussian noise, demonstrating broad applicability to scientific model discovery in biology, fluid dynamics, and material science.
Zitieren
APA
Kumar, S., Pandey, M., Lalita, Maheshwari, S., Kaur, N., & Chauhan, S. S. (2026). Data-Driven Discovery of Mathematical Models Using Artificial Intelligence. Computer and Decision Making: An International Journal, 3(6S), 949-956. https://doi.org/10.1922/comdem.v3i6s.18769
Vancouver
Kumar S, Pandey M, Lalita, Maheshwari S, Kaur N, Chauhan SS. Data-Driven Discovery of Mathematical Models Using Artificial Intelligence. Computer and Decision Making: An International Journal. 2026;3(6S):949-956. doi:10.1922/comdem.v3i6s.18769
BibTeX
@article{Kumar2026,
author = {Kumar, Sanjay and Pandey, Mohit and Lalita and Maheshwari, Shikha and Kaur, Navpreet and Chauhan, Surendra Singh},
title = {Data-Driven Discovery of Mathematical Models Using Artificial Intelligence},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {3},
number = {6S},
pages = {949--956},
doi = {10.1922/comdem.v3i6s.18769}
}
EUVIMED