The Research Of Public Opinion Diffusion And Control In Social Networks

In recent years,with the rapid development of social network platforms,social networks have become the main platform for obtaining and spreading information,which gradually replaces the traditional media.The generation and spreading of the information data in social networks is more convenient.It provides convenience for the spread of information,but at the same time provides a better platform for rumors and reactionary information.Different from the real society,public opinion of social networks cannot be restricted by time,space,which makes the public opinion of social networks more difficult to be controlled.It will become a wide influencing social event and endanger the security of the whole society if the development of negative public opinion cannot be controlled when it bursts.As a result,social network public opinion are paid more and more attention to by academia and industry.In this paper,the most representative social networking platform,Microblogging,as the researching object,being aimed to research the influence,prediction and control of the social network public opinion spread.Because the microblogging network has a large amount of data,serious information fragmentation,diverse interaction and fast information transmission characteristics,recent system audits or artificial real-time monitoring can't limit the spread of social network public opinion crisis information well.Therefore,for weibo social network public opinion acts as the main driving medium,how to automatically implement the impact assessment,prediction,control of microblogging public opinion and regulation in the spread of negative public opinion information,effectively to prevent the microblogging public opinion crisis and correctly guide the microblogging public opinion are the key problems of the social network public opinion security.Our paper mainly focuses on the perspective of public opinion security in socal networks,mainly conduct the research from the following aspects:First,for large-scale fragmented data and sparse relation data in microblogging,we propose an efficient co-clustering algorithm which processes sparse relation data of multi-entity.In order to take full advantage of multi-relational data when using this algorithm,we propose a robust constraint information embedding algorithm to construct relation matrix,and the performance of relation mining is improved by reducing matrix sparsity.In the sparse constraint block coordinate descent framework,relation matrix concurrently obtains cluster indication matrix of different entities by non-negative matrix tri-factorization.In non-negative matrix factorization,to ensure sparse structure of clustering result,a quick solution is achieved through efficient projection algorithm.Experiments on artificial and real data sets show that algorithm on three indicators has been improved obviously,especially the effect on extremely sparse data.Secondly,to measure the message diffusion influence of public opinion in socail networks,this paper proposes a directed tree model based on user interaction considering the history,type and frequency of interaction.User interaction matrix was used to describe the interactions between two users.A directed diffusion tree was generated from sparsification of interaction graph.This directed diffusion tree can be used to depict the fast diffusion process of information.The edges of directed diffusion tree were used to measure the information influence and identify the spam in microblogging.The experiments based on real the microblogging network data sets show that the tree model effectively depicts the message transmission process,while effectively measures the spread of news.Thirdly,we consider the effect of key users for public opinion diffusion and propose a dynamic linear model based on key users to predict message of public opinion in socail networks.In order to avoid the effect of structural attribute on predicting message diffusion,we portrayed the scale of message diffusion as the diffusion scale of multiple users,and build the diffusion effect function of users.To improve the efficiency of fitting function,we only consider key user to a built linear model.We use difference value between actual value and predicted value to detect key users and improve linear model based on new key users.The experiments based on real microblogging network data sets show that the model can effectively predict the messages propagation and find key users who can affect the performance of model prediction.Finally,the public opinion can not be effectively controlled by removing the relevant public opinion message.We propose an influence maximization for microblogging user group to lead public opinion.To reduce network scale,we delete useless user nodes,and construct a simplified microblogging network graph by ranking users.We employ the seed candidate set to construct a simplified microblogging network graph.In simplified microblogging network graph,We use influence greedy algorithm based on influence accumulation spread to find the seed set.In order to control the development of public opinion,we control the users of the seed set to post the lead information of public opinion.The experiments based on the real microblogging network data sets show that this method can be effectively applied to the public opinion control of a specific user community.