Social Mining In ZJU-SD Dataset

With the rapid development of Internet and ubiquitous computing, more and more social behaviors are recorded in the form of data, and collected into social dataset. A great number of social data attracts people from all walks of life, including sociologists, computer scientists, anthropologists and reseachers in the area of production, advertisement, marketing and so on. Researchers hope to explore social characteristics and regularity by mining these datasets from all kinds of views. In this thesis, we carry out one experiment to collect one social dataset which includes three-dimensional social information of users, and explore social characteristics of users by mining this dataset. The main work and contribution can be listed as follows:Social network datasets and social network analytical methods are first introduced in this thesis, and then the current research status and challenges. To deal with the challenge that most of current datasets are lack of diverse social data, we carry out one experiment to collect users’ three-dimensional social information:real-world mobility offline information, virtual-world on-line social information and self-report subjective information. Base on this dataset, we analyze a list of social behaviors of users from the view of users’encounter, movement and community. According to the social network topological graph, we define four social network graph:Mobile social network graph, online social network graph, friend social network graph and close friend social network graph. Following this, we explore the network structure of these four social net-work graphs, especially on the issue of network density,"small world", network assortative and network topological similarity. After that, this thesis studies the correlation between self-report social relationship and personality traits with the observed online and offline social data, and puts forward the prediction model to infer the self-report social relationship and personality traits. The conclusions and future work are depicted in the end of the dissertation.