Methodology for the automatic generation of error rules and adaptive adjustment of functions of pertenence through a network network architecture netfuz 1.0

Abstract

In the generation of the Fuzzy Inference Systems, the primordial task is the extraction and the tuning of the memberships functions and the fuzzy rules. However, when using the traditional methods to carry out this task, the obtained results are not the prospective ones and in most of cases serious inconveniences are presented. This article presents a methodological proposal base in Artificial Neural Networks that allows extracting the fuzzy rules and the parameters of the functions of membership of a Fuzzy Inference System type Sugeno automatically, leaving of a group of data input-output. The development of a software is contemplated that will facilitate the application in the control of processes, the prediction and the estimate of parameters.