Abstract: A method was proposed to evaluate the user’s attitude from the interaction experience and deduce the user’s behavior pattern by employing the interactive time series. Further, a new trust evaluation model was constructed by considering the following features: the user’s attitude, the interaction experience and the behavior pattern. The experiment results show that the presented model can be adapted to dynamic and complex social networks, and can efficiently quantify the influence on the trust of three kinds of information, and enhance the availability and accuracy of direct trust evaluation model in social networks.