Set membership prediction of nonlinear time series

A nonlinear prediction method based on a set membership approach is proposed. Such method does not need any assumption about the functional form of the model used for prediction, but uses only some information on its regularity. On the contrary, most of the existing prediction methods need the choice of a model structure and this choice is usually the result of heuristic searches. These searches may be quite time consuming, and lead only to approximate model structures, whose errors may be responsible of bad propagation of prediction errors, especially for the multi-step ahead prediction. Moreover, the method proposed in this paper assumes only that the noise is bounded, in contrast with statistical approaches which rely on assumptions such as stationarity, ergodicity, uncorrelation, type of distribution, etc. The effectiveness of the method is tested on simulated data and real word data (Wolf Sunspot numbers series), comparing the obtained prediction performances with those obtained by methods based on neural networks and on statistical models