Improving the Wang and Mendel’s Fuzzy Rule Learning Method by Inducing Cooperation Among Rules
Jorge Casillas, Óscar Cordón · 2000
Nowadays, Linguistic Modeling (LM) is considered to be one of the most important areas of application for Fuzzy Logic. It is accomplished by descriptive Fuzzy Rule-Based Systems (FRBSs), whose most interesting feature is the interpolative reasoning they develop. This characteristic plays a key role in the high performance of FRBSs and is a consequence of the cooperation among the fuzzy rules involved in the FRBS. A large quantity of automatic techniques has been proposed to generate these fuzzy rules from numerical data. One of the most interesting families of techniques, due to its simplicity and quickness, is the ad hoc datadriven methods. However, its main drawback is the cooperation among the rules which is not suitably considered. With the aim of facing up this drawback, which makes the obtained models not to be as accurate as desired, a new approach to improve the performance obtaining more cooperative rules is introduced in this paper. Following this approach, a concrete LM method based on one of the most known and widely used ad hoc data-driven methods, the Wang and Mendel’s method, is also presented. Its operation mode is composed of two stages: generation of the candidate rule set and combinatorial search of the rule set with best cooperation. Its behavior in the solving of two different applications will also be shown.