Abstract:
Understanding the dynamics of information diffusion in social networks contributes to a wide range of social studies. Among social networks, online social networks have drawn growing interest due to their richness, availability, and popularity nowadays. Such networks, which are often embedded in geographical space, have enabled information to spread at a relatively lower cost and higher speed and reach, compared to traditional ways of communication. This dissertation aims at exploring the spatiotemporal patterns of information diffusion in discussion about real-world events in online social networks, with special interest in geographical characteristics and representation. Specifically, this dissertation presents a methodology for studying and analyzing information diffusion in geographic space between sources and sinks of information. By doing so it highlights the information diffusion mechanisms that are in play at the intersection of the cyber and geographical environment, which can provide additional insights for higher-level decisions making. This dissertation also addresses the widely existing demand for traceable individual point information in data streams with geographical information, by designing an improved density-based stream clustering method. The method used not only meets the demand for finding cluster shapes, maintaining individual point information, and articulating point-cluster relationships, but also serves as the basis for spatiotemporal analysis and discovery of patterns hidden in the data.
