Recommendation Method Based On The Node Influence Of The Social Network

In the information age,the Internet provides users with a great deal of useful information to meet various needs of the users.However,we will meet a large amount of information,at the same time,users could not obtain the really useful and interesting information,which reduce the efficiency of information.The recommendation system could effectively provide users with more interesting and higher quality information,and enhance the user experience and user stickiness,reduce duplicate information which disgusts the users.As the Weibo network platform is widely used,the number of active users is increasing rapidly,microblogs recommendation technology has become one of the current research hotspots.At present,many recommendation algorithms of the Weibo platform ignore the problems that the inconsistence between the probability distributions of the microblog topics and the user interest probabilities and the poor quality and diversity of the microblog.In response to this,this thesis presents a recommendation method based on the influence of social network nodes.Its main research content:Firstly,in order to improve the quality of recommended microblogs and increase the diversity of Weibo,improving the MBUI-Rank influence algorithm of dynamic features method is needed,which takes the user influence and the influence of fans in the dynamic features into account.In the method of measuring user activity,the time factor is added so that users who' s activity is increased can be found.At the same time,improve the type of interest,official and self-checking users weight.For measuring the quality of microblogging,this thesis measures it from a single topic perspective into a multi-topic perspective,the purpose is to tap the variety of influential users,access to get high-quality and multi-type microblogs.Secondly,because of the SMLDA theme model ignores the influence of the comment on the theme,this thesis will optimize the theme extraction accuracy of Weibo based on the problem.The idea of optimization is to improve the accuracy of the topic extraction by combining the user interaction and blog comments when constructing the topic model.It aims to improve the problem of match the user interest distribution and the topic distribution of the microblog,which could to achieve the gold of accuracy recommend for the users who are interested.Finally,computing the interest of user with the Labeled-LDA by the user tags and history information,generating the recommendation list by computing the distribution of blogs and interests,which at last implement this thesis' s recommendation.It can improve the problem that the distribution of the topics of the recommended Weibo and the distribution of the users' interests are inconsistent,so as to improve the accuracy and diversity of the recommendation.Design experiments that,comparing the improved MBUI-Rank algorithm with PageRank and MBUI-Rank,it shows that improved MBUI-Rank has a higher coverage rate.At the same time,the improved SMLDA theme model is compared with SMLDA model and LDA model.The experimental results show that the improved SMLDA model is perplexity is lower.Based on the experimental results,the recommendation based on the influence of social network nodes in this thesis compared with the recommendation method based on other influence algorithms,his thesis' s recommendation of accuracy,recall and F1 are increased.