A Social Trust Model Considering Trustees' Influence

Online social networks can be viewed as multi-agent systems (MASs) built on top of social relationships. In these environments, relationships among agents are often formed through trust. Accurately estimating the degree of trust between agents is important in this case as social relationships are frequently leveraged to recommend products or services. Existing social trust models often utilize rating similarity between different agents to calculate how much they should trust each other. However, when a new truster enters the MAS and has not provided sufficient number of ratings, existing approaches cannot effectively advise the truster on which other agents can be trusted. To address this problem, we propose a social trust model that considers a trustee agent's influence in a social network. Evaluation based on the Epinions dataset shows that the proposed model significantly outperforms a state-of-the-art approach in social recommendation.