A methodology for constructing fuzzy rule-based classification systems
- Fernández Garrido, J. M.
- Requena Ramos, Ignacio
ISSN: 1134-5632
Año de publicación: 2000
Volumen: 7
Número: 2-3
Páginas: 185-197
Tipo: Artículo
Otras publicaciones en: Mathware & soft computing: The Magazine of the European Society for Fuzzy Logic and Technology
Resumen
In this paper, a methodology to obtain a set of fuzzy rules for classification systems is presented. The system is represented in a layered fuzzy network, in which the links from input to hidden nodes represents the antecedents of the rules, and the consequents are represented by links from hidden to output nodes. Specific genetic algorithms are used in two phases to extract the rules. In the first phase an initial version of the rules is extracted, and in second one, the labels are refined. The procedure is illustrated by applying it to two real-world classification problems.