Genetic Programming for the Identification of Nonlinear Input-Output Models

Linear-in-parameters models are quite widespread in process engineering, e.g., nonlinear additive autoregressive models, polynomial ARMA models, etc. This paper proposes a new method for the structure selection of these models. The method uses genetic programming to generate nonlinear inputoutput models of dynamical systems that are represented in a tree structure. The main idea of the paper is to apply the orthogonal least squares (OLS) algorithm to estimate the contribution of the branches of the tree to the accuracy of the model. This method results in more robust and interpretable models. The proposed approach has been implemented as a freely available MATLAB Toolbox, www.fmt.veim.hu/softcomp. The simulation results show that the developed tool provides an efficient and fast method for determining the order and structure for nonlinear input-output models.