Nonlinear Systems Modeling and Control Using Support Vector Machine Technique

This paper firstly provides an short introduction to least square support vector machine (LSSVM), a new class of kernel-based techniques introduced in statistical learning theory and structural risk minimization, then designs a training algorithm for LSSVM, and uses LSSVM to model and control nonlinear systems. Simulation experiments are performed and indicate that the proposed method provides satisfactory performance with excellent generalization property and achieves superior modeling performance to the conventional method based on neural networks, at same time achieves favourable control performance.