A Method Of Eliminating Redundant Rules And Its Application

The existence of redundant rules is one of the reasons of the generation of huge association rules.The redundant rules usually correspond with domain prior knowledge or have the same meaning with rules that have been mined. So the redundant rules present little new information and are meaningless for assisting decision-making. The existence of these meaningless rules not only causes the waste of mining resources, but also affectes the users'chosen of interesting rules and reduces the effectiveness of association results mining.Based on the related research, this paper summarizes and clarified the definition of redundant rules, analyzes the essential characteristics of domain prior knowledge which is used to eliminating redundant rules. And theorems and lemmas for judging redundant rules are proposed. To make this method can be applied in practice, based on these theorems and lemmas, a new redundant rule eleminating method is proposed, and the core algorithm is proposed and described with pseudo-code.This new method can avoid the generation of redundant rules before confidence computing.For the application and verification of this new method, an association rules mining prototype system combined with ontology search technology is designed and implemented. Ontology search function and redundant rules judging function are added to the existing association rules mining system and the meaningless rules can be eliminated more effectively.This paper analizes how the priori knowledge and the association rules with 100% confidence cause redundant rules, and attempts to study the method of eliminating redundant rules based on prior knowledge. Then a new method of eliminating redundant rules is supplied.This research comes from an actual project, and the research result is applied into practice. The method is proved to be effective in some preliminary practical application.