| Popis: |
Group recommender systems (GRSs) have played an important role in numerous online applications by providing recommendation to the group of users where satisfaction of the entire group is a major concern. In traditional GRSs, the relevance of all the groups and the items is considered equal which does not produce accurate recommendations. In this paper, we propose a formalization of the GRS based on the relevance concept using profile merging scheme where collaborative filtering (CF) is applied on each group profile to generate effective recommendations to the group by considering the ratings of the items, the relevance of the groups and the relevance of the items. Further, our GRS framework provides relevant similarity measures, relevant prediction and recommendation quality measures. The experimental results on the benchmark MovieLens dataset demonstrate the efficacy of our proposed GRS framework. |
