A New Method for Eliminating Redundant Association Rules

Abstract

As one of the fundamental data mining methods, the association rule mining has widely been used in many fields. However, the existence of massive redundant rules has made the analysis very difficult, and has been a main barrier to efficient utilization of discovered association rules. This paper studies the characteristic of commonsense knowledge which is use to eliminate redundant association rules, and analyzes the impact of commonsense knowledge and the special rules whose confidence is 100% on the generation of redundant rules. Some theorems and corollaries of redundant rules are proposed and a new method for eliminating redundant rules based on these theories is proposed. This new method can prune some redundant rules by using commonsense knowledge and the special rules without calculating confidence, so it improved the efficiency of mining and the utilization of discovered association rules.