Neuro-fuzzy based multi-step-ahead prediction

Neuro-fuzzy systems have been proposed for different applications for many years. In this paper, a neuro-fuzzy serial-propagated multi-step-ahead predictor is developed for time series prediction. The predictor consists of several individual neuro-fuzzy networks to produce a series of predicted values. Each network is trained by a hybrid learning algorithm. Two benchmark data sets are used to demonstrate the effectiveness of the proposed serial-propagated architecture. Experimental results show that our approach can provide more accurate predictions than other traditional methods.