Fuzzy wavelet Modeling using data clustering

In this paper, a novel approach for tuning the parameters of fuzzy wavelet systems which are used for modeling of nonlinear and complex systems is proposed. In fuzzy inference system, each fuzzy rule is analogous to a wavelet basis function multiplied by a coefficient. Using clustering techniques, the center of these basis functions are located in the detected center of clusters. In this way, not only the approximation accuracy is increased, but also the number of unknown parameters is decreased. The feasibility of the proposed method is shown by modeling two highly nonlinear functions. The comparison of the results using the proposed approach, with the previous schemes, shows the effectiveness and superiority of this algorithm.