Weibo Influence Measurement And Public Opinion Control Based On Multi-domain Division

With the rapid development of Internet technology and communication technology,a large number of social network platforms have appeared on the network,and Weibo is representative Due to the immediacy,autonomy and interactivity of Weibo,the deficiencies of traditional media can be improved and broken,so it has been greatly developed.It has become an important way for Internet users to publish and share information,and has gradually become popular.Internet public opinion platform.It is also because of the more free and varied way of speaking.The outbreak of sensational events often presents a virus-spreading type,which brings huge challenges to public opinion monitoringThis dissertation studies the issue of public opinion communication and control in different fields of social networks,and conducts research on three specific issues:domain division,user influence measurement,and public opinion control in different fieldsFirst,the method of user domain partitioning for Weibo users is studied.The domain division of users is mainly based on the microblogs they publish,and due to the characteristics of the Weibo platform,the published microblogs are short texts.Based on the research,this dissertation proposes a short text classification algorithm combining word vector,LDA topic model and CNN,which is based on the problem of short focus and feature sparsity.It is spliced with the theme vector trained by the LDA theme model and input into the CNN,which strengthens the connection between words and texts,thus achieving accurate and reliable short text classification,which lays a theoretical foundation for the follow-up work of this dissertationSecondly,the method of mining key users in Weibo is studied.Considering that in Weibo network,the dynamic change of attention caused by recent active users has an important impact on the influence of nodes,and the reciprocal edge in the network is more conducive to information transfer between users.A microblog network user influence ranking algorithm is proposed..In this algorithm,the weight of the microblog network is given,and the reciprocal edge is given higher weight.Compared with the dynamic change of the microblog concern relationship in a period of time,the dynamic incremental network is extracted,and the influence of the microblog network and the incremental network node are calculated respectively.And normalized processing,and finally get the influence order of Weibo users.By mining the actual data of Sina Weibo and using SIR model to verify the proposed algorithm,the results show that the proposed algorithm is better than other similar algorithms in improving the coverage and ranking prediction.At the same time,the incremental network has an important impact on node ordering.In addition,considering the factors of information forwarding tree,another user influence ranking algorithm is proposedFinally,the method of controlling public opinion for Weibo is studied.According to the proposed short text classification algorithm and user influence ranking algorithm,the proportion of each user in each field and the comprehensive influence of users are obtained respectively,and the influence of each field of the user is obtained through calculation,so that high influence in various fields can be found.Users,combined with the classic target immunization method,finally proposed a microblogging public opinion immunization strategy based on domain division.After experimental simulation analysis,it was found that the immune effect was improved compared with the traditional immunization strategy.