Centrality-based Group Formation in Group Recommender Systems

摘要: Recommender Systems have become an attractive field within the recent decade because they facilitate users' selection process limited time. Conventional recommender systems proposed numerous methods focusing on recommendations to individual users. Recently, due a significant increase in number of users, studies this shifted properly identify groups people with similar preferences and provide list for each group. Offering requires computational cost it is therefore often not efficient. So far, most impose four restrictive assumptions: (1) (2) groups, (3) average group members, (4) full knowledge network topological structure. To overcome these limitations, we propose novel approach which improves accuracy using centrality concept. In approach, central users are identified as heads then consequently formed. After formation, new profiling strategy provided aggregate members relative their centralities. Our evaluated different types compared several common strategies over MovieLens-1M dataset. Experimental results demonstrate that our formation profiling, based user measure, lead more accurate