Nonlinear system identification by SOM with GP based local modeling

Abstract

Genetic programming (GP) is a useful tool of nonlinear model building, however it tends to give much complicated model structure. Local modeling describes one nonlinear model as a set of sweet models, this approach is one of the practical ways to handle nonlinear systems. In this paper, a novel nonlinear system identification technique is proposed by combining GP and local modeling. In this approach, an identification procedure is divided into three steps, which are Self-organizing map (SOM) based clustering of the regression vectors consisting of observed input and output signals, local system identification by using GP, and model fusion of local models by fuzzy inference to provide the global model. The applicability of the proposed method is shown by the results of some numerical experiments.