A fuzzy association rule-based classifier for imbalanced classification problems

Imbalanced classification problems are attracting the attention of research community because they prevalent in real-world and impose extra difficulties for learning methods. Fuzzy rule-based systems have been applied to cope with these problems, mostly together sampling techniques. In this paper, we define a new fuzzy association classifier, named FARCI, tackle directly imbalanced problems. Our proposal belongs algorithm modification category, since it is constructed on basis state-of-the-art classifier FARC–HD. Specifically, modify its three stages, aiming at boosting number rules minority class as well simplifying them and, sake handling unequal rule lengths, also change matching degree computation, which key step inference process involved process. experimental study, analyze effectiveness each one components terms performance, F - score , base size. Moreover, show superiority method when compared versus FARC–HD alongside techniques, another approach, two cost-sensitive methods an ensemble.