Research On Prediction Technology Of Microblog Propagation Effect

As a new communication media, microblog has swept the whole world in short years.Compared with traditional media, microblog has attracted much attention because of its greatpropagation effect, at the same time challenges the information supervision. Propagation effect isthe sum of all impact and result that micro-blog causes to audience and society. The research onpropagation effect of microblog has very important significance to target marketing withmicroblog, sentiment supervision and guidance in microblog.Supported by the national863plan project, this paper researches on propagation effect ofmicro-blog, and aims to analyze and predict propagation effect quantitatively. Propagation effectis an abstract concept, which has different meanings and evaluation indicators in different media.Microblog is an information exchange network built by users following others. The messagepropagates continually through followers’ retweet behavior. Therefore, retweet scale andpropagation time are the two important indicators to reflect propagation effects. This paperpredicts propagation effect from the two aspects. The main contribute is as follows:1. Proposes a retweet scale prediction model based on retweet probability transition namedRSPM-RPT. To solve the problem of low accuracy in current retweet scale prediction methods,through the analyses on motivation of individual retweet behavior, a prediction model is builtwith five related features: publisher influence, acceptor activity, content importance, similaritybetween acceptor interestingness and content, intimacy between publisher and acceptor. On thatbasis, the RSP-RPT model is proposed with the analyses on propagation characteristic ofmicroblog, a method of retweet scale prediction is also given. The experiment with SinaWeibodata shows that the method proposed by this paper can predict different microb logs fromdifferent users with high accuracy.2. Proposes a propagation time prediction model based on RBF neural network namedPTPM-RBF. Because of the nonlinear and dynamic relationship between propagation time andfeatures of user and content, RBF neural network is introduced to fit the complex relationshipbetween them based on analyzing propagation time, and then the PTP-RBF model is proposed.The experiment with SinaWeibo data shows that PTP-RBF model has a quicker training speedand higher prediction accuracy than other prediction models.3. Designs a propagation effect analysis and prediction system based on SinaWeiboplatform. The system uses three-tier architecture, contains five modules: Certification andAuthorization Module, Data Collection Module, Data Storage Module, Data Process Module andPropagation Effect Prediction Module. The system is a distributed deployment, which canincrease efficiency of data collection and reduce the system cost. The experiment shows that the system can predict the microblog propagation effect in the two aspects of retweet scale andpropagation time after training with an amount of data.