Power of the Group Neighborhood in Memory-Based Group Recommender Systems

Recommender Systems play a significant role in helping users identify items worthwhile for them to consume. With the increase of adopting such systems a need for systems that help a group of users identify such items for the whole group to consume together has emerged. Early research has focused on strategies that combine individual preferences to generate group preferences without much consideration of the group context in the recommendation technique. In this paper, we explore neighborhood selection in the group context when applying a neighborhood-based Collaborative Filtering approach to recommendation. We identify several neighborhoods that are related to the group context and investigate their effect on recommendation accuracy when employing a neighborhood-based Collaborative Filtering. We evaluate the performance of such neighborhoods with respect to the group recommendation technique and the group size. We measure performance using a success@n measure. Results show improvements on the success rate of recommendations when identifying a neighborhood, for the group as a whole, rather than basing the recommendation on only the individual neighborhoods of the group members.