基于多特征的热门微博预测算法研究  

Research on the Prediction Algorithm for Sina Popular Micro Blog Based on Multi-features

作　　者：郑志蕴[1] 江国林[1] 张行进[1] 王振飞[1] 李钝[1] 

机构地区：[1]郑州大学信息工程学院

出　　处：《小型微型计算机系统》2017年第3期494-498,共5页Journal of Chinese Computer Systems

基　　金：郑州大学新媒体公共传播学科招标课题阶段性成果项目(XMTGGCBJSZ05)资助;河南省科技攻关项目(142102310531)资助;郑州市科技攻关计划项目(141PPTGG368)资助

摘　　要：随着微博的迅猛发展,微博舆情已经成为研究热点.以新浪微博为研究对象,分析热门微博的影响因素,提出一种基于多特征的热门微博预测算法.首先,对微博的原始特征进行分析,从中提取关键特征.其次,利用信息增益算法,根据微博的传播特征对微博的热度进行度量.最后,结合BP神经网络算法,根据微博的内容和博主特征,预测微博的传播特征,并由此推算微博的热度来预测该微博能否成为热门微博.实验表明,该算法的查准率可以达到75%以上,F1度量值保持在78%左右,能够对刚发布的微博进行热度预测,适用于微博营销和舆情引导等领域.With the rapid development of micro-blogs, micro-blog public opinion has become a research topic. Using sinamicro-blog as the proxy,this paper analyses the factors of popular micro-blog and proposes a prediction algorithm for popular micro-blog based on multi features. Firstly, this paper analyzes the original features of the micro-blog and extracts key features of the micro-biog. Secondly, according to the information gain algorithm, this paperuses the transmission feature of the micro-blog to measure the heat of themicro- biog. Finally, combined with the BP neural network algorithm, this paper usesthe content and blogger feature of the micro-blog to pre- dict the transmission feature of the micro-blog, with which we can calculate the heat of the micro-blog to predict that whether the micro-blog would be popular. It is indicated by experiments that,the precision of the algorithm can reach more than 75% ,the F1 metric can remain at around 78% ,and the algorithm can predict the micro-blog just been published, thus applied in the field of micro-blog marketing and the guidance of public opinion.

关 键 词：微博舆情 微博预测 信息增益 BP神经网络 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]