Skip to Main content Skip to Navigation

COCoFil2 : Un nouveau système de filtrage collaboratif basé sur le modèle des espaces de communautés

Abstract : Collaborative filtering traditionally exploits of user ratings on items, in order to connect people who share the same tastes and allow them to receive interesting recommendations from their peers. From our point of view, community management is an important issue because of the influence of community quality on recommendation performance. In this thesis, we have defined a model for multiple “community spaces”, and we have developed and evaluated it along the following lines.
Firstly, we have studied the nature of communities in the system, exploring what information can be exploited to build communities. In practice, users often receive interesting recommendations from friends, but also colleagues, neighbors, etc. Therefore, we have proposed to base the building of communities on multiple criteria in order to enrich the recommendations produced by the system.
Secondly, we have focused on the possibility for users to actually perceive communities as computed by the system: this possibility improves user acceptance of recommendations arising from these communities and is an incentive for users to contribute their own ratings into the system. We have proposed a process that produces 2D maps to allow users visualizing multi-criteria communities.
Finally, multi-criteria communities lead to two types of difficulties: first, on the user side, providing the information needed to be positioned in all of the community spaces, may require too much effort from users; second, on the system side, the computation complexity of some community spaces may lead to computation-intensive processes. Our community space model answers these problems by integrating tools to elaborate and apply appropriate strategies in the computation of communities as well as for automatic user positioning in multi-criteria communities. With the latter, we have shown that the recommendations produced in “cold-start” situations encountered by new users, are improved with the use of community spaces and appropriate automatic positioning strategies.
Document type :
Complete list of metadatas

Cited literature [117 references]  Display  Hide  Download
Contributor : An-Te Nguyen <>
Submitted on : Saturday, January 17, 2009 - 2:32:28 AM
Last modification on : Thursday, November 19, 2020 - 12:59:39 PM
Long-term archiving on: : Tuesday, June 8, 2010 - 8:35:53 PM


  • HAL Id : tel-00353945, version 1



An-Te Nguyen. COCoFil2 : Un nouveau système de filtrage collaboratif basé sur le modèle des espaces de communautés. Modélisation et simulation. Université Joseph-Fourier - Grenoble I, 2006. Français. ⟨tel-00353945⟩



Record views


Files downloads