The prevailing of Web 2.0 techniques has led to the boom of various online communities. Goodexamples are social communities such as Twitter, Facebook, Google+, and LinkedIn, which successfully facilitate the information creation, sharing, diffusion, and evolution among web users.As a result, a popular topic or event can spread much faster than in the Web 1.0 age. Indeed, whensearching for a recent popular event (e.g., Hurricane Irene or Toyota Recall) on Twitter, all theresults returned on the first page are created within the past five minutes.In such a scenario, the objective of my thesis is to advance the data mining technique to create a system that detects, tracks, and analyzes the evolution and diffusion of popular events in a social community. Specially, in the first part of the dissertation, I introduce a mining algorithm for popular event detection, which can efficiently and effectively extract widely adopted and meaningfulpatterns of user behaviors; in the second part, I depict a novel and principled probabilistic model to track the popularity index of events in a time-variant social community that consists of bothdynamic textual and structural information; in the third part of the dissertation, I address the problem of topic diffusions by studying the joint inference of topic diffusion and evolution in social communities, where contents and linkages in user-generated text information, together with socialnetwork structures, are used to facilitate the identification of topic adoption, the tracking of topic evolution, and the estimation of actual diffusion paths of any arbitrary topic.
PDF
download
878987797
028-85220240
