Multiple fuzzy neural networks modeling on sparse data based on a nonparametric regression technique

Combining neural networks and fuzzy systems is a great tool for modeling nonlinear systems. Few researches have presented useful or practical results on the case of lack of data, which does not provide necessary information for training the model. In this paper, we proposed a new modeling idea based on nonparametric regression, which provide us prior information for constructing the fuzzy system. Then a stable updating algorithm is proposed to train the membership functions. Due to the structure changes in the plant, a hysteresis switching algorithm is given to enable finite switch between the multiple fuzzy neural identifier.