Research On Web Page Recommendation Algorithm Based On Topic Model

With the development of social networks,Sina Weibo,as one of the largest social platforms,has brought tremendous pressure on information processing due to its vast user base and vast amount of posts and messages.Many valuable content and users are hidden behind Among the huge information.At present,there are new users on Sina Weibo,the influence of a single user on a subject is not easily measured,and potential friends are not easily found.In response to the above issues,this article will be through the following three sections,Sina microblogging users to make recommendations.First of all,because of the relationship between microblogging users and microblogging text divided into different user groups,and on this basis,using theme models to analyze the theme distribution of user blogs in different groups.In this paper,we mainly use the idea of k-means constrained clustering,on the basis of this idea,we improve the complex relationship between Weibo users and put forward the concept of tightness among users.At the same time,Select a target user's friends and their friends buddy clustering.Finally,a weakly constrained k-means algorithm based on the user-to-user tightness is proposed.Finally,the experimental comparison proves that the proposed algorithm improves the validity of the data on the social network.Secondly,according to the characteristics of blog posts and user characteristics of the user groups on Weibo,a comprehensive evaluation model of impact on Weibo user groups is proposed.This model combines the idea of ? TwitterRank algorithm,integrates the comments,forwarding,praise of blog posts and tag similarity between users to measure the influence of microblog user groups.At last,the experiment proves the rationality and validity of the influence model of this article,meanwhile it can evaluate the new users and the single users with the same theme.Finally,on the basis of influence,based on the theme similarity of the target users and the users in the most influential groups and testifying through the microblogging real data sets,the method of this article can better find new users and theme single users What can be recommended.Compared with the traditional recommendation method,the improved recommendation method has better accuracyand expansibility,and the actual recommendation content is more extensive.