Ranking Decision Rules Using Rough Set Theory—A Comparative Analysis

Abstract

Rule induction is one of the fundamental tasks in knowledge discovery system. The aim of rule mining is to discover hidden and interesting patterns from data contained in large datasets. Rough Set Theory provides efficient mathematical models suitable for decision rule induction especially in the case of inconsistent datasets. To develop measures for extracting significant rules automatically from a large number of rules generated by a data mining system is a challenging problem in rule generation. In this paper, decision rules produced by a conventional rule mining system are ranked based on a rule evaluation measure called Degree of Rule Significance. Based on this measure a rule ranking algorithm is designed to rank the generated decision rules. The ranking given by the algorithm is evaluated with the help of two rough set based rule evaluation approaches. Experimental results show the effectiveness of Degree of Rule Significance to identify important rules.