Approche hybride de recommandation des nouvelles collaborations.

With the spread of organizations and especially distributed ones, teams are highly dynamic and people are continually seeking for new collaborators. In this context, we propose an approach of collaboration recommendation that relies on community detection technique to find potential collaborators that can help in problem solving. Our contribution consists on proposing a combined metric that considers both social connections and homophily of participants. The proposed approach has been tested on a real graph representing a research laboratory where members represent researchers and interactions represent their relationships of co-publishing and co-supervision. MOTS-CLES : recommandation de collaborations, detection de communautes, interaction sociale,