Abstract
Genetic fuzzy rule selection has been successfully used to design accurate and compact fuzzy rule-based classifiers. In genetic fuzzy rule selection, first candidate rules are extracted by a data mining technique. Then only a small number of candidate rules are selected by a genetic algorithm to minimize a fitness function based on accuracy and complexity. In this paper, we incorporate the rule compatibility into the fitness function in order to obtain classifiers with clearer decision boundaries. We demonstrate the effectiveness of the proposed fitness functions through some simple examples.
