基于用户行为特征的微博谣言检测  

Microblog Rumor Detection Based on User Behavior Characteristics

作　　者：李艳君 张海军[1] 潘伟民[1] LI Yanjun;ZHANG Haijun;PAN Weimin(School of Computer Science Technology,Xinjiang Normal University,Urumqi 830054)

机构地区：[1]新疆师范大学计算机科学技术学院,乌鲁木齐830054

出　　处：《计算机与数字工程》2023年第4期850-854,共5页Computer & Digital Engineering

基　　金：国家自然科学基金-新疆联合基金重点项目(编号:U1703261);2019年度自治区创新环境(人才、基地)建设专项(人才专项计划--天山雪松计划)(编号:2019XS08)资助。

摘　　要：众多谣言在公开社交平台微博上肆意产生与传播,谣言检测有利于降低谣言对社会产生的不良影响。为探究微博用户的行为特征与该用户发布谣言的关联,提出一种基于用户行为特征的微博谣言检测算法(RDUC)。该模型主要以用户的点赞、转发和评论等行为特征作为主要参数,挖掘用户历史行为与谣言发布的关联,并且将ERNIE模型和DPCNN模型相结合对微博谣言事件进行检测。通过使用Ma公开数据集进行实验并与3种常用的谣言检测算法比较得出:该算法的准确率高达90.1%,高于这3种常用谣言检测算法。因此RDUC算法具有实际意义和应用价值。Numerous rumors are generated and spread wantonly on the public social platform Microblog.Rumor detection helps reduce the negative impact of rumors on society.In order to explore the relationship between Weibo user's behavior characteristics and their rumors,a microblog rumor detection algorithm based on user behavior characteristics(RDUC)is proposed.The model mainly uses the user's behavioral characteristics such as likes,reposts,and comments as the main parameters to mine the association between user's historical behavior and rumor publishing,and combines the ERNIE model and the DPCNN model to detect microblog rumor events.Through simulation experiments using the Ma public data set and comparison with three commonly used rumor detection algorithms,it is concluded that the accuracy of the algorithm is as high as 90.1%,which is higher than these three commonly used rumor detection algorithms.Therefore,the RDUC algorithm has practical significance and application value.

关 键 词：用户行为 ERNIE DPCNN 谣言检测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]