Saeid SOHEILY-KHAH - Dept. of Computer Engineering , Islamic Azad University
Mohammad Teshnehlab - Dept. of Electrical Enginnering, Khajeh Nasir Toosi University, Iran - Tehran

Nowadays, the usage of intelligent methods in predicting some quantities may be used as a powerful instrument for improvement of prediction models. In this paper, the objective has been the designing of a neuro-fuzzy network with a dynamic structure and effort has been made on .the basis of the number of input-output data pairs, to train the system via a rather improved method. In this direction, first at all, using the previously mentioned parameters, we define the fuzzy collection in such a manner to cover the input-output pairs. Then we setup the fuzzy regulations base and the fuzzy system and finally, through considering the threshold amount for the y in the conclusion section of the neuro-fuzzy network, we train the system.

Fuzzy System, Neural Network, Neuro-fuzzy network, Back Propagation, Forecasting