A Research On The Subject Interest Of Microblog Users Based On BTM Model

Because of its short,convenient,and instant features,Weibo has attracted a large number of users in recent years.Users have forwarded and commented on information on the Weibo social platform,which has enabled rapid dissemination of information,resulting in a large amount of information flow.In order to find user-interested information from massive information and achieve accurate delivery of microblog advertisements,we need to design a reasonable user interest model to mine user interest.Therefore,research on user interest acquisition is extremely important for the development of Weibo's website.This paper builds a user interest theme model based on BTM(Biterm Topic Model).The basic idea of BTM model is to search for "word pairs" in the entire corpus,which mainly aims at the establishment of the model of "word pair" generation process.Because the microblog text has the characteristics of less information and high-dimensional sparseness,it is not suitable to use LDA and other topic models to extract the interests of microblog users.As a word-pair model,BTM has a great advantage in short text information mining.This article takes Sina Weibo's user microblog text as a source of data and mainly does the following work:Firstly,this paper proposes a user topic interest extraction model based on BTM.It analyzes all the microblog users' microblog texts using this model,and obtains the user's interest topic distribution.Based on the vocabulary under each topic,the topic is summarized.User interest.Then,this article filters personal basic information,professional or professional,personal characteristics and other information from the user's personality tag,and retains user-related information as a reference for comparison with experimental results.Finally,this paper sets two variables of user activity(number of followers and number of fans)and microblog text,analyzes the effect of these two variables on the performance of BTM user interest topic model,and analyzes and explains experimental results's reason.Experiments show that the user activity has little impact on the performance of the BTM user topic model,and the number of microblog texts has a greater impact on it.The larger the number,the higher the accuracy of the BTM user topic model.At present,the interest mining of Weibo texts is mostly based on LDA.This paper conducts systematic experiments in BTM model to explore the performance of BTM model in Weibo short text user interest mining,and it is a BTM model in Weibo user interest mining.The application provides a reference.