Identification of chaotic process systems with least squares support vector machines

We investigate the nonlinear identification of chaotic process systems with least squares support vector machines, based on a case study of a parallel cubic autocatalytic reaction. State space reconstruction techniques are used to obtain a low-dimensional representation of the system in a different, but equivalent coordinate system. The performance of the support vector machine models are compared to corresponding models obtained when using multilayer perceptron neural networks, which are known to model chaotic dynamical systems well.