Information Diffusion Analysis And Model Construction In Microblog Network

Microblog is a social network platform for people sharing information andcommunication, which is based on users’ relationship. It plays a very important role inthe information distribution and public opinion propagation, with its convenient usage,rapid propagation and strong interaction. Microblog not only changes the form andstructure of information dissemination in network, but also changes the way and habitthat people access and dissemination information. Studying information diffusion inmicroblog will be helpful not only for correctly understanding its propagationcharacteristics and rule, designing information services which are coincident withusers’ habits, but also for monitoring, directing and dominating of microblog system.This thesis selects Sina Weibo as the main object of research, analyses topologycharacteristics of user network and characteristics of information diffusion, thendescribes information diffusion process with users’ behavioral characteristics inmiroblog and design information diffusion model based on users’ behavioralcharacteristics and simulates the evolution of information dissemination on mircoblogreal data. The main work of this thesis is presents as follows:At first, It statistics topological characteristics of the users’ network andinformation diffusion law. Though the sample data, analyses network topologicalcharacteristics and proves the impact of network structure on information diffusion byexperiments. Then we analyze characteristics of information diffusion, pathcharacteristics of information diffusion network and the factors that affect informationdiffusion network structure with reference of diffusion example.We describe information diffusion process according to analysis of users’behavioral characteristics, and design information diffusion model in microblog basedon users’ behavioral characteristics. According to the users’ location and role indiffusion process, we analyze the impact of poster, reader and disseminator oninformation diffusion. And with the history data of users’ post, repost information, thisthesis calculate the influence of poster, read habit of reader and repost preference ofreposter, as the model parameters. Then we call the model algorithm, simulatediffusion process on the real microblog data, and the simulation result demonstratesthat the model matches the real diffusion characteristics in microblog network. Thenwith the scope of the information diffusion as evaluation criteria, the experiments gethigher accuracy through spread scale forecasts on the micoblog subnetwork data andonline data in whole network. It provides theoretical basis for discovery and control ofpublic opinion in microblog network in future work.