Study On Ranking Method Of User Influence Based On Multi-attribute Decision Making

In the novel coronavirus pneumonia epidemic period,netizens are living in isolation,using entertainment,electricity providers,social networking and other application software for information acquisition,online shopping,online office,to accelerate users' habit of using the Internet.As the largest social networking platform in China,microblog's user scale and duration are far more than other social networking software.If the false negative network information is popular,it is easy to bring public opinion pressure to managers and challenges to social governance.So it is essential to identify high impact users to supervise public opinion effectively.This paper constructs the user influence evaluation system based on the three first level indicators and compares the common machine learning models of text classification and emotion analysis to determine the optimal way to quantify the indicators.And the user influence evaluation index system is combined with multi-attribute decision making discipline initiatively to build the user influence ranking model with different weight setting methods.This paper proposes INRank algorithm and INPRank algorithm with time span attribute based on OWA,and empirically analyzes the public opinion event data in entertainment,society and international fields.The main conclusions are as follows:(1)This paper uses multi-attribute decision-making method to solve the problem,which ignores the topic correlation in communication path.The verticality of microblog content and emotional tendency have influence on the change of user influence ranking.(2)The four algorithms proposed in this paper can be used for user influence ranking.OWA based on discrete normal distribution can reduce the influence of extreme value on the final ranking result to a certain extent.(3)The INRank model and INPRank model of interval number are introduced to fully consider the dynamic process of public opinion,which can more accurately and comprehensively rank the influence of users.This paper carries out differentiated public opinion control strategy for different types of users by subdividing the users' domain and emotional tendency.On the one hand,local immunization strategy is implemented for high influence users.On the other hand,the public opinion control strategy is implemented for users with different emotional tendencies,which includes publishing positive emotional information,personalized directional recommendation and isolating negative emotional users.