Trust Prediction in Online Social Networks

Online Social Networks (OSNs) have become an integral part of daily life in recent years. They have been used as a means for a rich variety of activities, such as seeking service providers or recommendations. In these activities, trust is one of the most important factors for participants’ decision-making process. Therefore, it is necessary and significant to predict the trust between two participants who have no direct interactions. My thesis aims to provide effective and efficient trust prediction approaches to evaluate trust values, which are introduced from the following four aspects. The first aspect of the work is to study the factors that affect trust in OSNs and solve the trust network extraction problem. OSNs contain important participants, the trust relations between participants, and the contexts in which participants interact with each other. All of such information has a significant influence on the prediction of the trust from a source participant to a target participant without direct interactions. In addition, the trust network, containing a truster and a trustee without direct interactions, is the foundation to perform trust prediction. The extraction of a small-scale trust subnetwork can deliver efficient and effective trust prediction results. We propose two heuristic algorithms called NBACA and NACA for the extraction of such subnetworks. The second aspect of the work is to address the trust prediction problem in the trust network without any contextual information. We first analyze and extract the features which affect the trust prediction from trust rating values in a trust network. Then, a new trust prediction model based on trust decomposition and matrix factorization is proposed to predict the trust value from a truster to a trustee. In this model, trust is first decomposed into trust tendency and tendency-reduced trust. Based on tendencyreduced trust ratings, matrix factorization with a regularization term is leveraged to