Group behavior diffusion model of social hotspots based on triadic structure and factor graphs. (15th May 2018)

• Record Type: Journal Article
• Authors: Li, Qian; Huang, Kai; Wu, Bin; Xiao, Yunpeng; Wang, Bai
• Is Part Of: Computational intelligence. Volume 34:Number 4(2018)
• Page Start: 1155
• Page End: 1177
• Publication Date: 2018-05-15
• Subjects: factor graph -- group dynamic -- information triad -- social network -- topic analysis; Artificial intelligence -- Periodicals; Computational linguistics -- Periodicals; 006.3
• DOI: 10.1111/coin.12176
• Languages: English
• ISSNs: 0824-7935

Abstract

Abstract: In view of the complex relationships involved in network structures, as well as the Markovian nature of information diffusion and the complexity of diffusion paths in social hotspots, we introduce and optimize a basic structure in social networks—the triad. A group behavior diffusion model is proposed based on the optimal triadic structure and factor graph. In addition to predicting individual participation behavior, the model forecasts group development trends of hotspots. First, considering the dynamics of social topic information diffusion, an information triad structure is proposed based on the traditional structure for analyzing the complex influencing factors. Second, in accordance with the Markovian nature of information diffusion, a group behavior diffusion model is proposed associated with the basic concepts and methods of random field theory. This model can not only capture the strengths of different influencing factors but also mine the dynamic patterns of topics. The experimental results reveal the efficiency of our model in both predicting individual participation behavior and perceiving group evolution trends in social hotspots.