TruGRC: Trust-Aware Group Recommendation with Virtual Coordinators

Abstract In recent years, an increase in group activities on websites has led to greater demand for highly-functional group recommender systems. The goal of group recommendation is to capture and distill the preferences of each group member into a single recommendation list that meets the needs of all group members. Existing aggregation functions perform well in harmonious and congruent scenarios, but tend not to generate satisfactory results when group members hold conflicting preferences. Moreover, most of current studies improve group recommendation only based on a single aggregation strategy and explicit trust information is still ignored in group recommender systems. Motivated by these concerns, this paper presents TruGRC, a novel Tru st-aware G roup R ecommendation method with virtual C oordinators, that combines two different aggregation strategies: result aggregation and profile aggregation. As each individual’s preferences are modeled, a virtual user is built as a coordinator to represent the profile aggregation strategy. This coordinator provides a global view of the preferences for all group members by interacting with each user to resolve conflicting preferences. Then, we also model the impact from group members to the virtual coordinator in accordance with personal social influence inferred by trust information on social networks. Group preferences can be easily generated by the average aggregation method under the effect of the virtual coordinator. Experimental results on two benchmark datasets with a range of different group sizes show that TruGRC method has significant improvements compared to other state-of-the-art methods.