The generation and the properties of the multi-step point predictors of the Self Exciting Threshold AutoRegressive Moving Average model has been examined. In the first part we focus the attention on the main properties of this class of models and on their different representations that can be derived extending, to the non linear case, some well know results given in Box and Jenkins (1976). Star...

This paper investigates the possibility of obtaining long-into-the-future reliable forecasts of observed nonlinear cyclical phenomena. Unsmoothed monthly sunspot numbers that are characteristically cyclical with nonlinear dynamics as well as their wavelet-transformed and wavelet-denoised series are forecasted through October 2008. The objective is to determine whether modelling wavelet-conversi...

Modern embedded platforms need sophisticated resource managers in order to utilize their heterogeneous computational resources efficiently. Furthermore, such are subject fluctuating workloads that unforeseeable at design time. Predicting the incoming workload could enhance efficiency of management this situation. But is case? And, if so, how substantial improvement? Does multiple-step-ahead pre...

An alternative approach, between much others, for mathematical representation of dynamics systems with complex or chaotic behaviour, is a radial basis function neural network using k-means for clustering and optimized by pseudo-inverse and particle swarm optimisation. This paper presents the implementation and study to identify a dynamic system, with nonlinear and chaotic behaviour, called Röss...

Abstract: Forecasting the future evolution of a system based only on past information comprises a central scientific problem. In this work we investigate the comparative performance of recurrent multi–layer perceptrons, trained through backpropagation through time and the differential evolution algorithm, to perform one–step–ahead predictions for the laser time series (Data set A) from the Sant...