Uncertainty-Informed Model Selection Method for Nonlinear System Identification and Interpretable Machine Learning - accepted version

Aug 4, 2026

Upload by Eliana

Uncertainty-informed model selection method addresses the challenge of choosing an optimal model among many candidates in data-driven modeling and machine learning. By explicitly modeling uncertainty, the approach tackles the fact that uncertainty ubiquitously arises during data modeling processes and makes model comparison difficult. Evaluation on data generated from a complex system model shows improved effectiveness versus conventional techniques, while requiring minimal training data length and limited assumptions on model types, enabling use across modeling frameworks.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages7

Reading Time18 min

AccessFree to download

Introduction

Motivation and challenge of model selection

Related work: information criteria and model selection methods

Scope and evaluation approach

This is a repository copy of Uncertainty-informed model selection method for nonlinear system identification and interpretable machine learning. White Rose Research Online URL for this paper: [https://eprints.whiterose.ac.uk/214151/](https://eprints.whiterose.ac.uk/214151/) [Version: Accepted Version](Version: Accepted Version) Proceedings Paper: Yuanlin, G. and Wei, [H.-L. orcid.org/0000-0002-4704-7346](H.-L. orcid.org/0000-0002-4704-7346) (2024) Uncertainty-informed model selection method for nonlinear system identification and interpretable machine learning. In: 2024 32nd Mediterranean Conference on Control and Automation (MED) . 2024 32nd Mediterranean Conference on Control and Automation (MED), 11-14 Jun 2024, Chania, Crete, Greece. Institute of Electrical and Electronics Engineers (IEEE) , pp. 909- 914. ISBN 9798350395457 [https://doi.org/10.1109/MED61351.2024.10566184](https://doi.org/10.1109/MED61351.2024.10566184) © 2024 The Authors. Except as otherwise noted, this author-accepted version of a paper published in 2024 32nd Mediterranean Conference on Control and Automation (MED) is made available via the University of Sheffield Research Publications and Copyright Policy under the terms of the Creative Commons Attribution 4.0 International License (CC-BY 4.0), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is properly cited. To view a copy of this licence, visit [http://creativecommons.org/licenses/by/4.0/](http://creativecommons.org/licenses/by/4.0/) Reuse This article is distributed under the terms of the Creative Commons Attribution (CC BY) licence. This licence allows you to distribute, remix, tweak, and build upon the work, even commercially, as long as you credit the authors for the original work. More information and the full terms of the licence here: [https://creativecommons.org/licenses/](https://creativecommons.org/licenses/) Takedown If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [eprints@whiterose.ac.uk](eprints@whiterose.ac.uk) including the URL of the record and the reason for the withdrawal request. [eprints@whiterose.ac.uk](eprints@whiterose.ac.uk)[ ](eprints@whiterose.ac.uk)[https://eprints.whiterose.ac.uk/](https://eprints.whiterose.ac.uk/) Final accepted manuscript 2024 32nd Mediterranean Conference on Control and Automation (MED) June 11-14, 2024 | Chania, Crete, Greece. pp. 909-914. Uncertainty-Informed Model Selection Method for Nonlinear System Identification and Interpretable Machine Learning* Yuanlin Gu, Hua-Liang Wei Abstract—Modeling uncertainty has been an active and important topic in the fields of data-driven modeling and machine learning. Uncertainty ubiquitously exists in any data modeling process, making it challenging to identify the optimal models among many potential candidates. This article proposesan uncertainty-informed method to address the model selection problem. The performance of the proposed method is evaluated on a dataset generated from a complex system model. The experimental results demonstrate the effectiveness of the proposed method and its superiority over conventional approaches. This method has minimal requirements for the length of training data and model types, making it applicable for various modeling frameworks. I. INTRODUCTION The typical process of data-driven modeling involves several stages including data collection, preprocessing, model training and model validation. For most data-driven modeling methods, it is often necessary to define some training parameters prior to model training. For instance, when building a neural network model, parameters such as the number of epochs, the estimation/optimization algorithm, evaluation metrics, and network structure (including layer types, number of layers, and neurons per layer) must be specified first [1], [2] . For some regression-based models, such as th

FAQs

What problem does the proposed method target?

It targets model selection—identifying the best model among many candidates when uncertainty affects data-driven modeling and machine learning.

How is the method evaluated?

The performance is evaluated using a dataset generated from a complex system model, followed by experiments comparing results with conventional approaches.

What practical advantages does the method provide?

It has minimal requirements for training data length and model types, making it applicable to various modeling frameworks.

Aug 4, 2026

Upload by Eliana