On Rule Generation Approaches for Genetic Fuzzy Systems

Marcos E. CintraMaria C. MonardHeloisa A. CamargoTrevor P. MartinAndrei Majidian

Genetic Fuzzy Systems have been researched for two decades and a considerable number of approaches have been proposed in the literature. De- pending on the strategy used by the genetic algorithm, the generation of candi- date rules is required to form the search space of the genetic algorithm. Specif- ically for the generation of these rules, proposals in the literature include the exhaustive generation of rules, the use of selection criteria over rules generated exhaustively, such as support, confidence, and degree of coverage, among oth- ers. This paper describes some of these methods and present their advantages and disadvantages in order to provide the reader with relevant information when deciding which method to use. A method for rule extraction with competitive advantages based on formal concept analysis is also proposed and preliminary results are discussed in detail. These results show evidence that our proposal is suitable for the task of forming a search space in terms of number of rules and processing time.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: