A modified community-level diffusion extraction in social network

Equipped with more convenient facilities and features, online social networks have become the most popular platform for people's communication. It is increasingly important to model information propagation in such networks. Most of the state-of-the-art algorithms of information diffusion model focus on individual-level diffusion and does not consider the impact of social relations on user's expression, making them either unable to uncover diffusion patterns accurately or unable to capture dynamically changing topics of text stream in social networks. To address these issues, we proposed a dynamic community-level diffusion model (DCDM) in this paper to capture diffusion patterns based on coordinated dynamic semantic analysis by multiple topic-word distribution and structure analysis. Comparative experiments are conducted on the real dataset from Tweet. Experimental results show our diffusion model outperforms the state-of-the-art methods.