Méthode de regroupement par graphe de voisinage

Fabrice Muhlenbach (1)

1 LHC - Laboratoire Hubert Curien

This work is related to the unsupervised machine learning problem. Some clustering methods, which are part of this research area, are based on the following hypothesis: the more two individuals are close in the representation space, the more they have a chance to belong in the same class. This paper presents a new clustering method that considers the proximity through the structure of a geometric neighbourhood graph.

Mots clés: fouille de données; classification non supervisée; graphe de voisinage

Domaines: Apprentissage [cs.LG]

Fabrice Muhlenbach. Méthode de regroupement par graphe de voisinage. Extraction et gestion des connaissances (EGC'2009), Jan 2009, Strasbourg, France. pp.361-366. ⟨hal-00364651v2⟩