Handling Item Similarity in Behavioral Patterns through General Pattern Mining

Abstract

Modeling human behavior on the Web is often performed by sequential pattern mining (SPM). However, the similarity between data elements often results in the decrease of the number of patterns mined. This work proposes to handle this similarity by managing multiple data sources representing different views of the data. We introduce G SPM, a behavioral pattern mining algorithm that takes advantage of multi-source data to handle the problem of data similarity. It adopts a selective mining strategy to limit the complexity and forms general patterns to limit the decrease of the patterns.Experimental results confirm that G SPM succeeds in handling the problem of item similarity. In addition, G SPM outperforms traditional approaches in terms of runtime and redundancy of the resulting set of patterns.