Teaching of neuro-fuzzy system on the basis of the method of difference areas

A new method of teaching MISO-fuzzy systems is reviewed and it uses in its structure fuzzy inference with linear membership functions. The peculiarity of this method is the usage of the difference of areas in the quality of defuzzification. The structural scheme of the worked out the neuro-fuzzy inference system was synthesized. The results of numerical modeling are shown and demonstrate the principle of the work of the suggested method and the comparative analysis with the traditional model of ANFIS is given. The worked out method increases accuracy of the fuzzy systems, and besides it is proved with the number of imitative experiments.