Core-Based Clustering Techniques

Abstract

Starting from model-based clustering simple techniques based on cores are proposed. A core is a dense region in the high-dimensional space that, for example, can be represented by its most typical observation, by its centroid or, more generally, by assigning weight functions to the observations. Well-known cluster analysis techniques like the partitional K-Means or the hierarchical Ward are useful for discovering partitions or hierarchies in the underlying data. Here these methods are generalised in two ways, firstly by using weighted observations and secondly by allowing different volumes of clusters. Then a more general K-Means approach based on pair-wise distances is recommended. Simulation studies are carried out in order to compare the new clustering techniques with the well-known ones. Moreover, a successful application is presented. Here the task is to discover clusters with quite different number of observations in a high-dimensional space.