Inverse system identification of nonlinear systems using LSSVM based on clustering

Abstract

In this paper we propose the algorithm of embedding fuzzy c-means (FCM) clustering in least square support vector machine (LSSVM). We adopt the method to identify the inverse system with immeasurable crucial variables and the inenarrable nonlinear character. In the course of identification, we construct the allied inverse system by the left inverse soft-sensing function and the right inverse system, and decide the number of clusters by a validity function, then utilize the proposed method to approach the nonlinear allied inverse system via offline training. Simulation experiments are performed and indicate that the proposed method is effective and provides satisfactory performance with excellent accuracy and low computational cost. © 2008 Springer-Verlag Berlin Heidelberg.