Modeling of nonlinear dynamic system using nu-support vector machines

The feed-forward neural networks trained with back-propagation learning algorithm have gained attention for the modeling of nonlinear dynamic systems. However, some inherent drawbacks like the multiple local minima problem, the choice of the number of hidden units and the danger of over fitting would make it difficult to put the networks into practice. This paper explores an additional possibility. We use nu-support vector machines for modeling of nonlinear dynamic system. The effectiveness of this new method is evaluated when the training data are noise-free as well as noisy. Simulation results reveal that the proposed method can obtain a good dynamic system model even when the training data are contaminated with additive noise of high levels.