Research On Social Influence Analysis For Microblogs

In recent years,the Internet is showing a rapid development trend;especially mobile Internet is increasingly affecting every aspect of people's lives.Sina Weibo gradually becomes one of the most important Chinese social platforms,because of its simplicity,terminal expansibility,and many other good user experiences.People can get information,enhance interpersonal relationships,participate in social issues and so on through Sina Weibo;and all of these activities bring huge amounts of social data.Sina Weibo is not only the platform of showing the public opinion and social activities,but also has a significant impact on people's business activities and even the country's security.Inferring key messages and influencers in micro-blogging services does do good to the online marketing and grasping the public opinion.This paper focuses on the problem of measuring the social influence of microblogs.This paper analyzes the factors that affect the social influence of microblogs,such as repost count,comment count,and like count;and designs a comprehensive evaluation method as a measurement of the social influence of microblogs.Then we verify the effectiveness of the comprehensive evaluation method.This paper focuses on the problem of predicating influential microblogs of micro-blogging services in China.Firstly,On the basis of analyzing the features provided by micro-blogging services,this paper develops a set of features that can be used to predicate the social influence of microblogs,such as word count of a single microblog,the sentiment of the microblog text and so on.In the follow-up experiment,the validity of the set of features is verified.Then we design and implement a method to predicate the social influence of microblogs based on Learning to Rank.Also the contribution of each feature to the prediction of microblogs' social influence is analyzed experimentally.Experimental results show that our method outperforms many other related methods,such as random Froest,Generalized Linear Model,and so on.