An Evolutionary Approach for Learning Rule Weights in Fuzzy Rule-based Classification Systems

Rule weights often have been used to improve the classification accuracy without changing the position of antecedent fuzzy sets. Recently, fuzzy versions of confidence and support merits from the field of data mining have been widely used for rules weighting in fuzzy rule based classifiers. This paper proposes an evolutionary approach for learning rule weights and uses more flexible equations, which are evolved by genetic network programming. We perform some experiments using 15 well-known UCI data sets to examine efficiency of these novel equations; then, to analysis the results obtained in this experiments, a nonparametric statistical test is used. The results show that the performance of the fuzzy rule-based classification systems can be improved by using this proposed method for rule weight specification.