Nonlinear dynamic system identification using least squares support vector machine regression

Abstract

The least squares support vector machine (LS-SVM) regression is presented for the purpose of nonlinear dynamic system identification. The LS-SVM achieves higher generalization performance than the multilayer perceptron (MLP) and radial basis function (RBF) neural networks and no number of hidden units has to be defined. Another key property is that unlike MLP training that requires nonlinear optimization with the danger of getting stuck into local minima. A difference with the RBF neural networks is that no center parameter vectors of the Gaussians have to be specified. The identification procedure is illustrated using simulated examples. The results indicate that this approach is effective even in the case of additive noise to the system. The LS-SVM can be used as an important alternative to MLP and RBF neural networks in nonlinear dynamic system identification.