自适应特征选择的图神经网络医保欺诈检测研究

打开文本图片集

中图分类号：TP391 文献标识码：A文章编号：1006-8228（2026）01-13-05

Abstract：Withthecontiuoustighteningofhealthcareinsurancefundsupervisionthereisstillanedtoaddressthechalengethat somefraudulentactivitiesarecharacterizedbyconcealmentandcomplexityExistingmodelsstillavecertainlimitationsinhandling classimbalanceandinterpretabilityandfurtheroptimizationorintegrationofnewmethodsisrequiredtoenhancetheir identificationcapabilities.Specifically，healthcareinsurancedatashouldbemodeledasheterogeneousdataandGraphAention Networks（GAT）shouldbeusedtofullyexplorethecontextualinformationofnodes.Additionally，theEvolutionaryFeature PurificationAlgorithm（FR）shouldbeintroducedtoadaptivelyselectfeatures，therebyimprovingthemodel'saccuracyand generalty.Thisapproachwillieldeteridentifationesultsmeettherasparencyquirementsofhealthcareinsuancditing and provide technical support for the healthcare insurance fund supervision.

Keywords：Medical InsuranceFraud;GraphNeuralNetwork；AdaptiveFeatureSelection;Medical InsuranceAuditing

0引言

医保基金事关每位公民的切身利益。（剩余7286字）