Multi-objective Frequent Termset Clustering

Large, high dimensional data spaces, are still a challenge for current data clustering methods. Frequent Termset (FTS) clustering is a technique developed to cope with these challenges. The basic idea is to first find frequent termsets and then to transform the resulting directed acyclic graph into a tree by deleting edges and termsets. While this technology was originally developed for document clustering, it can be applied in many other scenarios as well. Existing approaches to FTS clustering apply different heuristics to convert a set of frequent termsets into a final cluster set. In this work, we explore another approach. We first make the desirable properties of an FTS clustering explicit by stating different objective functions. We then show, how these functions are related to each other and that, in general, they are conflicting. This leads directly to the formulation of FTS clustering as a multi-objective optimization problem. We explore the ability of this approach to pr...

• Type: Conference
• Year: 2007
• Where: LWA
• Authors: Andreas Kaspari, Michael Wurst
• Keywords: Data Clustering Methods, Document Clustering, Frequent Termsets, LWA 2007, Software Engineering