Mining Potential Opinion Leaders Method Based On Topic Evolution

With the advent of the web2.0 era and the rapid development of the Internet,traditional media has been gradually replaced by new media generated in the Internet era.Through the Internet,information spreads faster and in more ways,making it difficult for users to choose and make decisions in the face of multifarious information on the Internet.Therefore,users tend to choose the information provided by opinion leaders.Opinion leaders play an extremely important role in the process of information dissemination and public opinion evolution.Nowadays,many scholars judge the actual role of opinion leaders by calculating their influence.The research adds the subject and the influence of the respondents view for the user to the model,the influence of using the text similarity measure between respondents between transition probability,with the comment text emotional tendency transition probability measuring reviewers to respondents,in order to transfer matrix structure,according to the PageRank algorithm to get the user's influence score after iteration.Then,by establishing the DTM topic evolution model,I cut the timelines of each topic into eleven parts according to the time stamp and the trend of the topic's heat.According to the established time window,the text information of the topic is processed,and the topic distribution of the topic text under the time window is judged.Then the topic distribution of the answer is judged by the answer of the respondent.By judging the difference between the two distributions,the opinion leaders who conform to the trend of topic evolution are identified,and the opinion leaders found are evaluated.The experimental results show that the improved algorithm based on PageRank can identify opinion leaders more accurately,cover more people,wider topics,and fit the topic more closely.Moreover,DTM model can be used to select opinion leaders who are more in line with the trend of topic evolution.