A Python library for nonlinear system identification using Multi-Gene Genetic Programming algorithm

TL;DR

This paper introduces a Python toolbox that uses Multi-Gene Genetic Programming for structure selection in nonlinear system identification, enabling efficient modeling of complex systems with customizable evaluation functions.

Contribution

The paper presents a novel Python library implementing MGGP for nonlinear model structure selection, integrating parameter estimation, simulation, validation, and user-defined evaluation functions.

Findings

Effective structure selection for nonlinear models

Supports gray-box identification with prior knowledge

Facilitates data-driven modeling in various fields

Abstract

Models can be built directly from input and output data trough a process known as system identification. The Nonlinear AutoRegressive with eXogenous inputs (NARMAX) models are among the most used mathematical representations in the area and has many successful applications on data-driven modeling in different fields. Such models become extremely large when they have high degree of non-linearity and long-term dependencies. Hence, a structure selection process must be performed to make them parsimonious. In the present paper, it is introduced a toolbox in Python that performs the structure selection process using the evolutionary algorithm named Multi-Gene Genetic Programming (MGGP). The toolbox encapsulates basic tools for parameter estimation, simulation and validation, and it allows the users to customize their evaluation function including prior knowledge and constraints in the individual structure (gray-box identification).