Cluster Characterization through a Representativity Measure

Clustering is an unsupervised learning task which provides a decomposition of a dataset into subgroups that summarize the initial base and give information about its structure. We propose to enrich this result by a numerical coefficient that describes the cluster representativity and indicates the extent to which they are characteristic of the whole dataset. It is defined for a specific clustering algorithm, called Outlier Preserving Clustering Algorithm, opca, which detects clusters associated with major trends but also with marginal behaviors, in order to offer a complete description of the inital dataset. The proposed representativity measure exploits the iterative process of opca to compute the typicality of each identified cluster.