A fuzzy rule-based system for ensembling classification systems

Tomoharu Nakashima, Gaku Nakai, Hisao Ishibuchi

Abstract: In this paper, we propose an ensembling method for pattern classification problems. The characteristic feature of our ensembling method is that two different types of fuzzy rule-based systems are used. One is a fuzzy rule-based classification system that suggests a class of an input pattern. The other is a fuzzy rule-based ensembling system that assigns a weight to the suggested class by each classification system. Our ensembling method consists of one fuzzy rule-based ensembling system, several fuzzy rule-based classification systems, and a gating node that finally determines the final classification of the input pattern. Computer simulations show the effectiveness of our ensembling method.