Research On The Influential Users Discovery Method Based On The Microblog Community

With the rapid development and improvement of the Internet,Sina microblog,as a representative social platform in the Web2.0 has achieved rapid development.It has become a new way for people to communicate and obtain information.People can use these ways to share the drops in life and publish their views on events.Users can pay attention to their favorite users.In the process of information dissemination,users' views or comments with many followers can profoundly affect their followers' views and determine the trend of public opinions.Therefore,it has been found that influential users in the microblog community have become the focus of research in this area of research.With the continuous development of microblog,microblog has appeared in a kind of malicious false users who pays attention to other users or specializes in marketing,we call it "spammer" users.Due to the existence of such users,the user's influence cannot be calculated correctly,and it makes microblog become a distribution center for rumors,causing microblog credit crisis and affecting social stability.In response to the above issues,the research work in this paper mainly includes the following points:This paper proposes an advertising spammers detecting model(ASDM)that combines multiple features of microblog text topic features,users attribute features and behavior features.The model first analyzes the difference between "spammer" users and normal random users based on the users' behavioral features.The behaviors of normal random users posting,reposting and commenting are decided based on the users' hobbies or time.The "spammer" users exhibit certain regularities in their behavioral features,such as regularly posting,reposting,commenting,etc,and sending a large number of repeated or similar microblogs within a certain time interval.Then the model takes into account the difference between the topics distribution of "spammer" and the normal random users in microblog texts.The topic model LDA is used to obtain the topic probability distribution of the users' microblog,and the repetition degree evaluation function defined in this paper is used to calculate the repeatability of the user's two adjacent microblog texts.Finally,we evaluated the effect of the user of the forwarded microblog source on the degree of text duplication.Experimental results show that the detection model proposed in this paper is more effective.The above research has weakened the false influence brought by the "spammer" in the microblog data and calculates the influence of users on this basis.At present,research on user influence usually only considers user attributes,interaction information,or network structure,and calculates user influence from one or several aspects.Due to the text sparseness of the microblog,less research has been conducted on the use of microblog topics to measure user influence.However,influential users often have greater influence in certain fields and may have weaker influence in other fields.To solve this problem,this paper proposes a topic preferences supervised random walks(TP-SRW)algorithm based on user topic preferences to discover influential users.The algorithm considers the microblog users' own attributes and the microblog text topic features,constructs user-link relationships based on user interaction information,and combines the similarity of the interest of the microblog topics that represents the homogeneity of the users,thereby constructing directed weighted networks.Then we will detection influential users according to the ranking algorithm proposed in this paper.Finally,in order to verify the effectiveness of the proposed model and algorithm,relevant experiments and analysis were performed on the microblog dataset.The experimental results show that the model proposed in this paper can identify "spammer" more effectively.And in terms of calculating users influence,the algorithm of this paper is better than users of Twitter Rank,Influence Rank and Leader Rank which also considers topical factors in different time periods.The algorithm of this paper is more effective in mining the topics community.