Trust inference in complex trust-oriented social networks

Many social networking platforms have emerged on the Web, such as MySpace and Facebook. In those networks, most participants are usually physically unknown in real life and have no prior direct interactions with each other. Hence, it is necessary to form complex trust-oriented social networks with more trust related information and infer trust values between participants. The inferred trust results provide important indications for some applications, such as introducing products to trustworthy people or recruiting trustworthy employees from social networks. Traditional trust inference mechanisms are based on simple trust networks and can hardly deliver realistic trust results. In this paper, we first propose a complex trust-oriented social network structure containing trust values, social relations and recommendation roles. We also propose a novel Bayesian network based trust inference mechanism taking trust values, social relations and recommendation roles into account. Experiments demonstrate that our proposed trust inference mechanism performs well in complex trust-oriented social networks.