Genetic Programming for System Identification

Linear in parameter models are quite widespread in process engineering, e.g. NAARX, polynomial ARMA models, etc. This paper proposes a new method for nonlinear structure selection for linear in parameter models. The method uses Genetic Programming (GP) to generate nonlinear input-output models represented in tree structure. The main idea of the paper is to apply Orthogonal Least Squares algorithm (OLS) to estimate the contribution of the branches of tree to the accuracy of the model. The proposed method speeds up the convergence of the GP and results in more robust and interpretable models. The simulation results show that the proposed tool provides an efficient and fast method to determine the order and the structure of nonlinear input-output models.