Information Networks: Problems, Theories and Applications

Nowadays, online information networks widely exist in our daily lives. The flourishing of various types of information networks have raised numerous challenging problems, as well as innovative applications. This dissertation introduces our latest research progress on several information network related problems, the theories behind our methodologies, as well as innovative applications related to information networks. In the first part of the dissertation, we focus on one specific problem: network inference problem. The problem emerges from the scenario when we need to uncover a hidden information network from the temporal records of disseminated messages. In order to achieve more accurate inference, we propose a general framework to fuse the preference of users and the distributions of diffused messages to uncover topic-sensitive information networks. Moreover, we also propose a clustering- embedded framework and embedding-based approaches to improve the efficiency of inferring information networks. Moreover, we introduce one of our models built on information networks, which has a successful impact on real-life business activities. To be more specific, in this application, we discuss how to leverage the abundant information contained in online social networks to improve the effectiveness of offline sales. Although the model is designed for the task of offline sales, it is general enough to be used in other similar business applications.