基于贝叶斯网络的社交网络用户影响力研究  

Research on Social Network User Influence Based on Bayesian Network

作　　者：吕永旭 LV Yong-xu(School of Statistics and Data Science,Nanjing Audit University,Nanjing 211815,China)

机构地区：[1]南京审计大学统计与数据科学学院,江苏南京211815

出　　处：《电脑与信息技术》2024年第6期81-86,共6页Computer and Information Technology

基　　金：国家自然科学基金(项目编号:72074117)。

摘　　要：针对社交网络中日渐泛滥的虚假信息,提出了一种基于贝叶斯网络的社交网络用户影响力模型。首先,为处理假新闻的选择偏差,排除用户与谣言间未观察到的负向互动,提出利用贝叶斯网络与逆倾向评分结合来学习用户与假新闻间正向互动的无偏假新闻分享行为。其次,选取并计算与用户影响力相关特征,用户间的关系和用户自身属性,建立社交网络用户影响力模型。与有偏假新闻相比,学习无偏分享的贝叶斯网络对谣言检测准确率、召回率等更高,且所得用户影响力排名,与社交平台算法较为接近。与其他传统算法相比,该模型的分类误差更趋于稳定,实用性得到了提高,并分析了影响力排名的影响因素。In response to the increasing proliferation of false information in social networks,this paper proposes a social network user influence model based on Bayesian networks.First,in order to deal with the selection bias of fake news and eliminate unobserved negative interactions between users and rumors,it is proposed to use Bayesian network combined with inverse propensity scoring to learn unbiased fake news sharing with positive interactions between users and fake news.Behavior.Secondly,select and calculate the characteristics related to user influence,the relationship between users and the user's own attributes,and establish a social network user influence model.Compared with biased fake news,the Bayesian network that learns unbiased sharing has higher rumor detection accuracy and recall rate,and the obtained user influence ranking is closer to the social platform algorithm.Compared with other traditional algorithms,the classification error of this model is more stable,the practicability is improved,and the influencing factors of the influence ranking are analyzed.

关 键 词：贝叶斯网络 虚假信息检测 社交网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]