Research On Group User Profiling Of Microblog Based On Biterm Topic Model

With the continuous development of information technology and the Internet,social media platforms,such as microblog and WeChat,have developed rapidly,and gradually have a tremendous impact on people's daily life.In the process of freely communicating,expressing personal emotions and sharing personal opinions on these platforms,people generate a large amount of information and data.How to effectively utilize these massive information and create more valuable user services is the focus of current research.Taking full account of the characteristics of micro-blogging data,this thesis aims to propose a novel GUPTM(Group User Profile Model based Topic Model)based on Biterm Topic Model.It closely links user's background information,social information and subject interest,and fosters the group user profiling of microblog,systematically,comprehensively and carefully.At the same time,the application of group user profiles in network public opinion governance is studied in detail,and finally new countermeasures are provided for network public opinion governance.The main contents of this thesis are as follows:Firstly,a new clustering method of microblog group users based on Biterm Topic Model is proposed.Obtaining the information freely released and shared by users from the microblog platform,establishing the model data set,analyzing the user relationship and preprocessing the data,we use the Biterm Topic Model to mine the user's topic interest.We calculate user's information background similarity,user's social information similarity and user's topic interest similarity respectively,and then blend these similarities to build a more effective similarity calculation method.Based on the calculation results of this method,the K-means clustering is used to realize user clustering.Secondly,a group user profile model of microblog based on Biterm Topic Model is constructed.We elaborate the framework and process of the group user profile model.And the parameters of GUPTM model are reasoned by using Gibbs sampling algorithm after pretreatment of microblog data to obtaine the key parameters of the model.The new group user clustering method proposed in this thesis is used to cluster user groups,and then the group features of microblog users are extracted by using the clustering results.At last,the group user profiling on microblog is constructed by using the group characteristicsFinally,under the background of network public opinion governance,the research on the group user profiling of microblog is carried out.The research results show that the group user profiling of microblog can effectively help guide public opinion and improve the efficiency of public opinion governance.Correspondingly,this thesis puts new countermeasures for network public opinion governance,providing some theoretical reference for follow-up researchers.