The Research On Informaiton Diffusion Modeling Over Social Networks

In our world today, how to make full use of connections between people to make the information propagate rapidly and efficiently is of great value both theoretically and practically, so influence maximization and community mining are two hotspots in current social network analysis, but these problems both depend on a specific diffusion model. Construct a diffusion model that can render the real-world phenomena accurately is the target of our research.Multiple information diffusion processes are widely taken place in social networks, such as spreading knowledge of multiple products that of same kind through word-of-mouth recommendations, but few works have done to study the dynamic theory of multi-information diffusion presently. Focusing on this situation, in this paper, a new multiple information competitive diffusion model is proposed based on the classic LT(Linear Threshold)model, in which the differences between individuals combined with the structure of the social network have been considered, which is an important factor that has been ignored in past researches, the important parameter bu ,vhas been improved and a new calculating formula of influence is given, which has been proven worked in simulations.In addition, few works has been done to study the diffusion mechanism in the on-line social network, of which is different from traditional social networks, so in this paper, a new information diffusion model based on on-line social networks is proposed, in which the influence of the information source is introduced. Simulation results show that the model, having the same characteristics with online social networks, contributes to a more profound understanding of information dissemination behavior on online social network.Finally, a new multi-information competitive diffusion model oriented to Micro-blog is proposed based on the combination of the two models that introduced before. Through empirical studies on the activity of users on the Sina Micro-blog, we find that the threshold value in the model follows a power-law distribution, while how to assign a proper value to the threshold parameter is an unsolved problem that obsesses people for a long time.