Generalization of c-means for identifying non-disjoint clusters with overlap regulation

Chiheb-Eddine Ben N'Cir (1) , Guillaume Cleuziou (2, 3) , Nadia Essoussi (1)

1 LARODEC - Laboratoire de Recherche Opérationnelle de Décision et de Contrôle de Processus
2 Equipe Hultech - Laboratoire GREYC - UMR6072
3 LIFO - Laboratoire d'Informatique Fondamentale d'Orléans

Clustering is an unsupervised learning method that enables to fit structures in unlabeled data sets. Detecting overlapping structures is a specific challenge involving its own theoretical issues but offering relevant solutions for many application domains. This paper presents generalizations of the c-means algorithm allowing the parametrization of the overlap sizes. Two regulation principles are introduced, that aim to control the overlap shapes and sizes as regard to the number and the dispersal of the cluster concerned. The experiments performed on real world datasets show the efficiency of the proposed principles and especially the ability of the second one to build reliable overlaps with an easy tuning and whatever the requirement on the number of clusters.

Keywords: Multi-label data; Clustering evaluation; Overlap regulation; Overlapping clustering

Domains: Machine Learning [cs.LG]

Chiheb-Eddine Ben N'Cir, Guillaume Cleuziou, Nadia Essoussi. Generalization of c-means for identifying non-disjoint clusters with overlap regulation. Pattern Recognition Letters, 2014, 45C, pp.92-98. ⟨10.1016/j.patrec.2014.03.007⟩. ⟨hal-00978269⟩