An Attribute Reduction Algorithm Based on the Maximum Dependency and Minimum Redundancy of Attribute

Abstract

The classical attribute reduction algorithms based on attribute dependence in rough set theory only select attributes which have a larger degree of dependence on decision attribute and don't consider attribute redundancy. This paper points out that only selecting condition attributes with a large degree of dependence on decision attribute is not enough, the redundancy between condition attributes should also be taken into account. In allusion to this matter, an algorithm based on the maximum dependency and minimum redundancy of attribute is presented. The results of experiments which are carried out on the UCI data sets suggest that the presented algorithm has gained favorable results.