Weibo User Forwarding Behavior Prediction Research

With the development of Internet technology and the popularity of 4G mobile network,WIFI network and the intelligent mobile terminal,more and more Internet users get and release information via social media.In China,microblog has gradually become the main platform for people to get information and interact with each other.In recent years,the research of information dissemination and user behavior has been widely studied by scholars both at home and abroad.The current researches on the reposting behavior focus on the three aspect,including the motivation of reposting behavior and its influence factors based on the Theory of Reasoned Action,and the topology and characteristics of microblog reposting network based on the theory of information transmission,and the user characteristic and the prediction of user behavior based on the machine learning algorithm.Few researcher consider that whether the user's interests may trigger his reposting behavior,and how much the text semantic information may effect on the user's reposting behavior.In this paper,the text semantic information of a microblog is fully considered for this problem,and the process of microblog user's receiving information,browsing information,making information behavior is analyzed.The influence factors of receiving user behavior lie in user's attitude and subjective norm,and then the research hypothesis is proposed based on the Theory of Reasoned Action,which contains the characteristic of the releasing user,the characteristic of the receiving user,the interaction between releasing user and receiving user,the semantic topic of microblog text,and the data form.The topic probability distribution of microblog text and user interest could be obtained based on the LDA topic model,and then the semantic similarity of microblog text and user interest could be calculated.The prediction model is trained based on binary logistic regression model,then the prediction experiment is conducted based on crawled data and the result is evaluated.This study found that semantic similarity of microblog text and user interest,the reposting activity of the receiving user the interaction degree between the receiving user and the releasing user have a significant affection on the reposting behavior.While,the impact of releasing user and the data form of the microblog not significant to the reposting behavior.The accuracy rate of the prediction model which is based on the validated influence factors is more than 90%,and the rate of missed diagnosis and misdiagnosis are both less than 10%.In conclusion,the model proposed in this paper can predict the behavior of users more accurately,which be applied in marketing prediction,public opinion monitoring and other fields as reference.