Research On Information Propagation In Social Networks

With the development of information technology,online social networks have become a powerful tool of information dissemination,which promotes rapid and large-scale dissemination of information and other content.Development of social networks has undergone rapid change,sina microblog and wechat are recent domestic popular online social tools.A huge data,which is produced by hundreds of millions of users' release,comment,reply,retransmission and other actions,diffuse through network rapidly,so that different people in different areas with different interest could receive various kinds of information easily.Information sharing has been well implemented.Therefore,the study of news dissemination mechanisms has got important practical significance.With respect to the information diffusion probability model,there are six factors this paper takes into consideration to study the information dissemination probabilistic model.The six factors are the information disseminator feature,the information receiver feature,the relationship between disseminator and receiver,the information characteristics,the history information feature of disseminator and receiver.This paper use Bayesian logistic regression method to conduct the training and parameter estimating based on sina microblog dataset.This paper establishes an information dissemination probabilistic model that is in line with the realities of the sina microblog network.The establishment of the information dissemination probabilistic model is beneficial to combine independent cascade model,linear threshold model and the competitive structure of the model to create a more realistic model of information dissemination.In terms of the research in influence maximization,this paper firstly describe the classic influence maximization problem,and introduce the corresponding solution algorithms and measurement.Then,combining with the reality of the spread phenomenon,the adaptive influence maximization problem,not specifying a select set,adaptively selecting the combination of spread seeds to achieve the purpose of maximizing influence,is proposed.Considering the phenomenon of competition for news dissemination,this paper conducted a study to maximize com--petition for influence spread at the same time.This paper presents the extreme value of the M-based algorithm based on influence earnings,compared with the traditional Monte Carlo greedy algorithm.With sina mircoblog's user data,this paper carried out the simulation experiments,which maximizes the influence of multi-information competitive diffusion with selected seed node,to verify the efficiency of M-based algorithm to a certain extent.At last,this paper carried out the statistical analysis of the degree distribution and the average distance characteristics of the seed nodes,which gives suggestions for the selection of reference seed node.