Fast and Simple Deterministic Seeding of KMeans for Text Document Clustering

Abstract : KMeans is one of the most popular document clustering algorithms. It is usually initialized by random seeds that can drastically impact the final algorithm performance. There exists many random or order-sensitive methods that try to properly initialize KMeans but their problem is that their result is non-deterministic and unrepeatable. Thus KMeans needs to be initialized several times to get a better result, which is a time-consuming operation. In this paper, we introduce a novel deter-AQ1 ministic seeding method for KMeans that is specifically designed for text document clustering. Due to its simplicity, it is fast and can be scaled to large datasets. Experimental results on several real-world datasets demonstrate that the proposed method has overall better performance compared to several deterministic, random, or order-sensitive methods in terms of clustering quality and runtime.
Liste complète des métadonnées

Littérature citée [22 références]  Voir  Masquer  Télécharger
Contributeur : Julien Velcin <>
Soumis le : mercredi 12 décembre 2018 - 22:52:48
Dernière modification le : mercredi 20 novembre 2019 - 02:42:30
Archivage à long terme le : mercredi 13 mars 2019 - 16:28:53


Fichiers produits par l'(les) auteur(s)




Ehsan Sherkat, Julien Velcin, Evangelos E. Milios. Fast and Simple Deterministic Seeding of KMeans for Text Document Clustering. 9th Conference and Labs of the Evaluation Forum (CLEF), Sep 2018, Avignon, France. pp.76-88, ⟨10.1007/978-3-319-98932-7_7⟩. ⟨hal-01953432⟩



Consultations de la notice


Téléchargements de fichiers