融合动态传播网络与双重特征差异的社交网络谣言检测模型

打开文本图片集

关键词：特征差异；动态传播网络；谣言检测；社交网络

中图分类号：TP391 文献标志码：A 文章编号：1001-3695（2025）10-015-3012-07

doi：10.19734/j.issn.1001-3695.2025.03.0061

Rumor detection model in social networks integrating dynamic propagation networks and dual feature differences

Xu Guiqiongat，He Sihua，Li Weiminb （a.SchoolofManagement，b.SchoolofComputerEngineering&Science，ShanghaiUniversity，Shanghai2OO44，China）

Abstract：Mostof theexistingrumordetectionmodelsrelyheavilyonstaticinformation，makingitdificulttoreflectthedynamicpropagationcharacteristicsofrumors.Moreover，these modelsoftenoverlookthefeaturediferencesanddynamicevolutionin dimensionssuch as emotionalpolarityand thematicsemantics between theoriginal postsandcomments.Tosolve this problem， this paperinnovativelyproposedanintegratingdynamicnetworkanddualfeature diferences model（DNDF）forrumordetection insocial networks.Themodelaimedtoimprovedetectionefectivenessbyanalyzingtheevolutionofmultidimensionalfeatures. Firstly，itusedadualfeaturediffrencemoduletoanalyzefeaturevariationsinemotional polarityandthematicsemantics betweenoriginalpostsandcommentsequences.Thenitcombined propagation graphsequencesandapplied BiSTMtogeneratedifferential featuresequences.Finally，itintroducedaco-atentionmechanismtosrengthentheinteractivelearningbetweentext features andemotionaldiference features，aswellasbetween thematicdiffrence features.Experimentsonpublicdatasets PHEME and WEIBO show that the DNDF model increases the accuracyby 0.3% and 2% respectively.The model outperforms mainstreambaseline models inmultiple indicators，such as F1 ，andconfirms its effectivenessin rumor detection in social networks. Key words：feature difference；dynamic propagation network；rumor detection；social network

0 引言

社交网络特有的开放性与便捷性，使其演变为网络谣言传播扩散的主要途径。（剩余19248字）