New a fuzzy inference system and ensemble method wi̇th parametri̇c t-norm operator for ti̇me seri̇es predi̇cti̇on

The thesis consists of two main issues. First issue is that Parametric Hamacher T-norm operator is applied to Adaptive Wavelet Networks (AWN) and Adaptive Neuro-Fuzzy Inference System (ANFIS). Product and min. T-norm operators are usually performed in Adaptive Neuro-Fuzzy Inference System. The major feature of parametric Hamacher T-norm is to provide several T-norm operator by means of its parametric structures. In proposed methods while Gaussian membership function is utilized for Adaptive Neuro-Fuzzy Inference System's condition part, wavelet function is used for Adaptive Wavelet Networks for the same manner. First order polynomial functions are available in consequence parts of both these models. Proposed model is tried chaotic time series prediction and system identification problems. Furthermore, optimum values of unknown parameter are determined by Broyden-Fletcher-Goldfarb-Shanno algorithm based on approximate Newton method. Moreover, proposed model is compared with other studies in the literature. The other proposed model is to apply ensemble method on the first proposed model. In the last phase of Ensemble model weighted average technique is applied. In order to determine optimal weights, learning algorithm is implemented and weighted average is calculated. The second proposed model is experimented on financial time series. Ensemble Adaptive Neuro-Fuzzy Inference System and Ensemble Adaptive Wavelet Networks is compared with each other.