An Extension to Hierarchical Conceptual Grouping Distance Based

Abstract

Hierarchical Distance-based Conceptual Clustering (HDCC) is a general approach to conceptual clustering.
HDCC extends the traditional distance-based agglomerative algorithm by producing on the fly conceptual descriptions of the discovered clusters.
One of the main contributions of HDCC is its theoretical framework, which provides a set of mathematical tools and theoretical results useful for the analysis of consistency between distances and generalisation operators in the context of HDCC’s algorithm. The framework defines three levels of consistency based on the divergences between the clustering hierarchies induced by the linkage distance and the new hierarchies of concepts and clusters induced by HDCC’s algorithm.
Inspired by the concept of distance-based generalisation proposed by Estruch (2008), in this work we revise and compare the sufficient conditions for distance-based generalisation operators vs. the properties defined in HDCC and
we extend the framework by adding a new level of consistency –the level of distance-based dendrograms.}