社交关系超图网络增强的多图融合兴趣点推荐方法

打开文本图片集

中图分类号：TP391 文献标志码：A 文章编号：1001-3695（2026）04-019-1129-10

doi：10.19734/j. issn.1001-3695.2025.08.0278

Multi-graph fusion with enhanced social relation hypergraph network for Pol recommendation

Cheng Shulin†，Yu Zhenqiang，Xie Wenqing，Jiang Jing （School of Computer Science and Information，Anqing Normal University，Anqing Anhui 246133，China）

Abstract：Asforthlimitationsofexistingapproaches inefectivelyextractingsocialrelationsandintegatingthem withspatio temporal information，thispaper proposedmulti-graph fusion with enhanced socialrelation hypergraph networkforPolrecommendation（MSRHN），which fuses local spatio-temporal enhanced graphs，global interaction hypergraphs，andsocialrelation hypergraphs for preciserecommendation.The local spatio-temporal enhanced graph model constructedaspatio-temporal enhanced graph neural network basedonuser-Pol interactions，employing an asymmetric propagation mechanism to modelspatio temporaldependenciesandacuratelyextractlocalspatio-temporalinformation.Theglobal interactionhypergraphmodelbuilt aninteractionhypergraphneuralnetworkbasedonigh-orderuser-Polinteractions，capturingigh-ordercollborativesignals throughatwo-steppropagation mechanismtoefectivelymodel global interaction patterns.Thesocialrelation hypergraph model constructedasocialrelationhypergraphneuralnetworkbasedonsocialconnectionsbetweenusers，and integratedinteraction informationamong multipleuserstoprofoundlyexplorehigh-ordersocialrelationships.Finally，thethreegraph modelswere fusedtogeneraterecommendationresults.Experimentsonthreepublicdatasets demonstrate thatthe MSRHN methodoutperforms existing baseline methods.

Key words：point-of-interest（Pol）recommendation；spatio-temporalenhanced graph；socialhypergraph network；multigraph fusion

0引言

基于位置的社交网络（location-based socialnetwork，LBSN）作为一种特殊的社交网络，为用户构建了一种开放的交互平台。（剩余27650字）