Role-Based Contextual Recommendation

In this paper, we present a role-based contextual recommendation approach, containing a role mining algorithm and a role-based recommendation algorithm. Specifically, role represents common preference or behavior pattern among a group of users, which can be mined from context information and profiles. Role expresses user´s context-aware interests, preferences, and requirements. Roles are then used to infer trust relations between users, which can be applied to achieve high recommendation quality in contextual recommendation. Experiments on real dataset show that our approach outperforms state-of-art recommendation approaches.