面向图神经网络的多层网络节点分类研究综述

打开文本图片集

关键词：多层网络；图神经网络；复杂网络；节点分类

中图分类号：TP393.02 文献标志码：A 文章编号：1001-3695（2026）04-002-0972-13

doi：10.19734/j. issn.1001-3695.2025.09.0309

Overview on node classification in multilayer networks based on graph neural network

ChenKexin，DingCangfeng†，Zhu Ye，Cao Bohao （College ofMathematics andComputer Science，Yan'an University，Yan’anShaanxi 716ooo，China）

Abstract：MultilayernetworknodeclasificationisacriticaltaskincomplexnetworkresearchItaimstoidentifyandcassify nodes withinanetwork touncovertheir intrinsic multidimensional structureandsemanticcharacteristics.GNN，saframework capableofdirectly process graph-structureddata，canefectivelycapturetheunderlying paterns ofnodesandedges，aswellas more profound semantic features，through message-passing mechanisms.In node classfication tasks， GNN exploitsgraph structure informationandnode featuresbyagregating information fromneighboring nodes and propagating messages.Comparedtotraditionalnode classfication methods，GNNlearnsnodefeaturerepresentationsadaptively，reducing relianceon manual featureengineering and enhancing clasificationaccuracyThis papercategorizedandreviewedrecent developments in GNN-basedmultilayernetworknodeclassificationmethods.Firstly，itoutlinedthecoreconceptsofmultilayernetworksand GNN，and provideddefinitionsandcharacteristicsofvariousmultilayernetworkmodels.Then，itcomprehensivelysummarized thelatestadvancementsinGNN-basedmultilayer networknodeclasificationmethods，andcategorized them intothreeparadigmsbasedonthelearningapproach：semi-supervised，unsupervised，andself-supervisedlearing.Thispaperalsoanalyzed the performanceof these methods in downstream tasks，such as social networks，academic networks，and E-commerce platforms.Finallyitprovidedathoroughsummaryofexistingesearch，iscussdthelimitationsofurrntmetods，adsge ted potential directions for future research.

Key Words：multilayer network；graph neural network（GNN）；complex network；node classification

0 引言

现代复杂网络科学为理解复杂系统带来了重大进展。（剩余39156字）