Research On Weibo User Recommendation Based On Topic Model

In recent years,Weibo,as a social application,has developed rapidly with the development of mobile client technology.The information in Weibo has also grown exponentially,which raises an issue of information overload.It is a challenge to search the content which we are interested in the “information ocean”.In response to the above issues,this paper proposes a user-focused recommendation based on the topic model.The specific research contents are as follows.Firstly,through the research on the topic model LDA,it is found that although the LDA-based topic model can obtain the user's Weibo topic probability distribution,and can calculate the user's topic similarity according to the topic probability distribution,on the emotionally rich social platform,it is slightly inadequate for the model ignores the user's emotional factors.This paper proposes an RJS algorithm for calculating user similarity.The idea of RJS algorithm is to get the user's theme probability distribution and sentiment tendency through the JST model.The similarity of the user's topic probability distribution is calculated by using the JS distance,and the final user similarity is calculated based on the user's sentiment orientation to obtain the user's similarity score.Secondly,the Weibo user influence model UIM is proposed based on the characteristics of Weibo.The UIM model divides the user influence into three parts: Weibo influence,user activity,and the number of users' fans.Through qualitative analysis of these three indicators,the calculation method of the user influence score is obtained;the weights of the three indicators were calculated using the analytic hierarchy process and the results were tested;and the three indicators were linearly combined to obtain the user influence score.The calculation formula can be used to calculate the user's influence rating.Finally,a new recommendation algorithm named R-UIM is proposed based on the combination of RJS algorithm and UIM model.The R-UIM algorithm is used to get the comprehensive score and recommend the highest rating to the target user.The experimental results show that the recommendation algorithm proposed in this paper has better performance in Weibo attention recommendation compared with the standard LDA,LDA+UIM and RJS models.