Nonlinear System Identification using ARX and SVM with Advanced PSO

Autoregressive with exogenous variable (ARX) model and least square support vector machine (LS-SVM) regression were used the field of linear and nonlinear system modeling. The structure of linear ARX model is concerned with the number of time lags of the model and coefficients. While the determination of the system order is very important in the performance of the estimated model, the automatic selection method is still not established. In addition, the parameters of LS-SVM effects on the performance of model absolutely. In this paper, the improved particle swarm optimization (PSO) is introduced as the technique of selecting the number of time lags in these problem. Furthermore, the optimum parameter values of LS-SVM model, weighting factor and kernel parameter, are obtained more quickly by the proposed PSO algorithm. Through a simulation example, the effectiveness of the proposed schemes is illustrated.