基于马尔科夫随机场的微博用户转发行为预测  

Predicting Microblog User Forwarding Behaviors Based on Markov Random Fields

作　　者：王宁[1] 高光 柴争义[1] WANG Ning;GAO Guang;CHAI Zhengyi(School of Computer Science and Technology,Zhoukou Normal University,Zhoukou,H enan 466000,China;School of Network Engineering,Zhoukou Normal University,Zhoukou,Henan 466000,China)

机构地区：[1]周口师范学院计算科学与技术学院,河南周口466000 [2]周口师范学院网络工程学院,河南周口466000

出　　处：《中文信息学报》2018年第6期107-113,共7页Journal of Chinese Information Processing

基　　金：国家自然科学基金(U1504613;41401463);河南省高校科技创新团队计划(171RTSTHN009);河南省知识产权局软科学研究项目(20170106041);周口师范学院青年基金(zknuB3201601;zknuB315204)

摘　　要：微博用户转发行为预测是微博社交网络消息扩散模型构建的基础,在图书阅读推广、舆情监控与市场营销等领域有着广泛的应用。为了提高用户转发行为预测的精度,该文在马尔科夫随机场框架下综合分析了用户属性与微博内容特征、用户转发行为约束等因素对用户转发行为的影响,并在逻辑回归模型的基础上构造了相应的能量函数对用户转发行为进行了全局性的预测。实验结果表明,微博用户转发行为不仅取决于用户属性、微博内容等特征,而且也受到与其相邻用户转发行为的约束。相对于传统算法该文算法可以更准确地对用户转发行为进行建模,因而可获得更好的预测结果。Weibo users＇ forwarding behavior prediction is the foundation of weibo social network message diffusion model, which can be widely applied in book reading promotion, public opinion monitoring, marketing management and other fields. Under the framework of Markov random field, this paper comprehensively analyzes the effects of user attributes,the weibo contents,and the user forwarding behavior constraints. The logistic regression model is em- ployed to construct the prediction model of the user＇s forwarding behavior. The experimental results show that the forwarding behavior of weibo users depends on the user attributes, the micro-blog contents, as well as the forwarding behavior of their neighboring users. Compared with the traditional algorithms, the proposed algorithm can model the user＇s forwarding behavior more accurately.

关 键 词：新浪微博 转发预测 能量优化 逻辑回归 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]