Extracting Non-Redundant Episode Rules Based on Frequent Closed Episodes and Their Generators

Abstract

Aiming at discovering causal relationships between frequent episodes,episode rule mining has been broadly applied in many fields such as sensor data processing,network security monitoring,finance securities managing,transaction log analyzing,and so on.To mine the non-redundant episode rules from an event sequence,an algorithm called Extractor is proposed in this paper.Extractor discovers all frequent closed episodes and their generators by employing the support definition of both minimal and non-overlapping occurrences and the depth-first search strategy,which assures the quality and efficiency of mining frequent closed episodes and their generators.Moreover,Extractor avoids redundant generator checking by utilizing the Apriori Property of non-generators.In addition,Extractor generates non-redundant episode rules directly from frequent closed episodes and their generators,which improves the quality and efficiency of generating episode rules.Experiments have proved the effectiveness of the proposed method.