Geração de Regras de Decisão Fuzzy Utilizando a Teoria dos Rough Sets

Abstract

This paper proposes a new method to automatically generate fuzzy rules based on rough sets teory and fuzzy sets teory. The derived rules are concise with respect to the number of antecedent terms and present high coverage rate. The classier system based on these fuzzy rules was tailored to discriminate between possibility of classication and impossibility classication. The proposal was tested with ve public databases provided by University of Wisconsin: the Iris, the Wine, the Wisconsin Diagnosis Breast Cancer considering 10 attributes (Wdbc with 10 attributes), the Wisconsin Diagnosis Breast Cancer considering 30 attributes (Wdbc with 30 attributes) and the Wisconsin Prognostic Breast Cancer (Wpbc). The classication accuracies obtained for Iris, Wine, Wdbc with 10 attributes, Wdbc with 30 attributes and the Wpbc were 100%, 100%, 99,03%, 98,03 e 93,90%, respectively.