Information Diffusion Mechanisms in Online Social Networks

A great deal of research interest has arisen in the area of information diffusion since the emergence of online social networks (OSN). Many models for describing how information diffuses in OSN have been proposed. But so far, the mechanisms of information diffusion remain largely unexplored. We try to find the mechanisms from the perspective of both network structure (micro-scale level) and popularity of information (macro-scale level). We focus on how these two levels interact with each other impacting information diffusion. Based on a Twitter data set which contains 196 million tweets, 10 million users, and 3 million hashtags, we perform a temporal analysis by calculating the time-evolving properties of network structure, including node degree, global clustering coefficient, the number of the nodes in the largest cluster. We then investigate how popularity evolves over time as the network structural properties change. We finally find that the moment when hashtag enter into bursting period is relevant to the largest cluster of users discussing the hashtag. When largest cluster changes from a closely connected small cluster to an emanative large cluster, hashtag enters into bursting period. We find this moment by calculating global clustering coefficient and the number of the nodes in the largest cluster. Our study on the interactive mechanisms between micro-and macro-scale levels reveals the essence of information diffusion and provides a necessary theoretical basis for predicting popularity or public sentiment warning.