| With the booming of online social network, it provides a rich data source for analyzing the relationship between people. For example, how people connect; how people's behavior can be influenced by neighbors, and how information broadcasts through social networks. By studying the structure and functionality of social network, it gives a new prospective on human behavior model, social organization model and information propagation model. From the view of sociology and game theory, a person's decision-making and behavior are greatly affected by their social groups, i.e., family, social class, job category, etc. In the work, we focus on studying how social network plays the important role on people's behavior and various of social network models, which try to give the explanation on how do social networks form.;In the dissertation, we present four effective methods to analyze the social network structure and its related applications. First, we introduce a joint personal and social latent factor (PSLF) model which combines the collaborative filtering and social network modeling approaches for social recommendation. Second, we develop a new social influence model, referred to as Socialized Gaussian Process (SGP) for socialized human behavior modeling. Third, we present a general Bayesian framework for co-occurrence friends modeling in social networks. Fourth, we develop a nonparametric ranking model (NRM) for link recommendation tasks. |
