Research On User Profiling Of Microblog User Based On Dynamic Topic Model

Recently,social networking services,such as Facebook,Twitter and SinaWeibo,have developed rapidly,and the number of its users has increased year by year.Social media is not only an important way of entertainment in daily life,but also a platform for brand building,precision marketing and other personalized services for enterprises.The personal information and behavior information left by users on the social media has become an important source of data for enterprises to study user characteristics,understand user needs,and deeply analyze users.Through the analysis of user data,user profiling technology can express users as multi-label user prototypes,help enterprises accurately locate target users,which is the basis of personalized recommendation services and has important commercial value.In this context,this paper mainly studies the construction of microblog user profiling based on dynamic topic model.In the traditional social user profiling construction,the researchers regard the text content published by users as an unordered corpus,and then extract the user's interest characteristics.However,the user's interest characteristics will change over time,and the recent feature has a higher weight,which is more conducive to describe the user characteristics accurately.Therefore,this paper introduces the time factor into the interest extraction process,and analyzes the changes of user interest in the time dimension through the dynamic topic model.Overall,the main content and innovations of this article can be summarized in two aspects:1.A Weibo user interest extraction method based on dynamic theme model is proposed.According to the timestamp of microblog content,the corpus is divided into different time slices to realize the evolution of the theme over time,and finally obtains the user's interest characteristics.2.A sub-dicided portrait of different types of Weibo users is constructed,which divides the user into four different kinds according to the user's activity index.Besides,this paper constructs a more comprehensive user profiling based on both static characteristics and interest characteristics of the user,and analyzes the differences between different portraits of different type of users.