DI-UMONS : Dépôt institutionnel de l’université de Mons

Recherche transversale
(titres de publication, de périodique et noms de colloque inclus)
2017-01-25 - Article/Dans un journal avec peer-review - Anglais - 9 page(s)

Fremal Sébastien , Lecron Fabian , "Weighting Strategies for a Recommender System Using Item Clustering Based on Genres" in Expert Systems with Applications, 77, 105-113

  • Edition : Pergamon Press - An Imprint of Elsevier Science
  • Codes CREF : Technologies de l'information et de la communication (TIC) (DI4730), Informatique appliquée logiciel (DI2570)
  • Unités de recherche UMONS : Management de l'Innovation Technologique (F113), Informatique (F114)
  • Instituts UMONS : Institut de Recherche en Technologies de l’Information et Sciences de l’Informatique (InforTech)

Abstract(s) :

(Anglais) Recommender Systems are effective to identify items that could interest clients on e-commerce web sites or predict evaluations that people could give to items such as movies. In this context, clustering can be used to improve predictions or to reduce computational time. In this paper, we present a clustering approach based on item metadata informations. Evaluations are clustered according to item genre. As items can have several genres, evaluations can be placed in several clusters. Each cluster provides its own rating prediction and weighting strategies are then used to combine these results in one evaluation. Coupled with an existing collaborative filtering recommender system and applied on Yahoo! and MovieLens datasets, our method improves the MAE between 0.3 and 1.8%, and the RMSE between 4.7 and 9.8%.

Mots-clés :
  • (Anglais) Recommender System
  • (Anglais) Clustering
  • (Anglais) Weighting Strategies