Study on method of on-line identification for complex nonlinear dynamic system based on SVM

Support vector machine is a learning technique based on the structural risk minimization principle, and it is also a kind of regression method with good generalization ability. This paper analyses the disadvantage of the nonlinear dynamical systems identification method based on neural networks, and presents an online support vector machine method to model nonlinear dynamical systems. Theoretical analysis and simulation result indicate that this method has the merits of high learning speed, good generalization as well as approximation ability, and little dependence on samples set. The present method has the better prediction precision than that of the approach based on the neural network.