Graph-Based Hierarchical Conceptual Clustering in Structural Databases
Cluster analysis has been studied and developed in many areas for a wide variety of applications. The purpose of applying clustering to a database is to gain better understanding of the data, in many cases through revealing hierarchical topologies. The authors are working on extending the Subdue structural knowledge discovery system with clustering functionalities. Past works related to ours are an incremental approach called Cobweb [Fisher 1987], and its extension, Labyrinth [Thompson & Langley 1991], that can represent structured objects using a probabilistic model.