Based On The Hierarchical Theme Model To Carry Out The Opinion Leader Mining Research In The Social Network

Since the birth of Web2.0,the amount of information in the Internet showed rapid growth,the information quality issues also followed,the most effective means of obtaining valuable information in the complex social network is to find opinion leaders on the Internet who have rich expertise and information,and these people are willing to spread knowledge and ideas.However,the current social network of opinion leaders mining research has these following problems:1.Most of the views about leaders mining research is based on the global perspective,thus ignoring the limitations of the field of opinion leaders and the limitations of expertise.2.Most of the opinion leaders mining research is based on the network topology,while ignoring the content of the opinion leaders,the opinion leader's personal attribute information and interactive information between users.3.Although many scholars have considered these characteristics of opinion leaders,but most of them just simply do multi-angle analysis,and did not integrate these into a system for a comprehensive study.In order to solve these problems,this paper chooses the knowledge-sharing community zhihu as the background and the research object,the hierarchical theme model HDP is introduced into the text mining process to get the theme,and then the specific topic area is selected.And after the comprehensive analysis of the user influence index system,on the basis of it,combined the index system with the PageRank algorithm.The zhihu-leadear-rank algorithm is presented in this paper,which further improve the mining of opinion leaders.This paper mainly carried out the following research work:First of all,this paper introduced the hierarchical probability model HDP in the process of user's text theme mining,discovered the potential topic through HDP.Experiments shows that HDP can better find potential topics in users' text information than LDA.Then,according to the characteristics of opinion leaders,this paper choose the user's attributes,the interactive information and relationship information between users to develop the index system of user influence.And then,this paper established the user influence index system,and combined it with PageRank algorithm.Then this paper presents the zhihu-leadear-rank algorithm,it has solved the problem of the influence of nodes in the original PageRank.Finally,some information is crawled based on the crawler,which includes user-related personal attribute information and user interaction information in a specific subject area,then the zhihu-leadear-rank algorithm is introduced to carry out the user influence ranking calculation.Finally,the information is integrated,the information includes the user's personal attribute information,users' interaction information,users'relationship structure information and the users'published text content information.Thus conducting a study of opinion leaders in specific areas.This article has overcome some of the shortcomings of predecessors in the process of mining the views of leaders.