福建师范大学学报（自然科学版）2026，Vol.42Issue(1)：1-9,9.DOI:10.12046/j.issn.1000-5277.2024070021
一种基于图同构时空网络的交通流预测模型
A Graph Isomorphism Spatio-Temporal Network for Traffic Flow Prediction
摘要
Abstract
Accurate traffic flow prediction is crucial for the effective operation of intelligent transportation systems.To address this need,this paper proposes a graph isomorphism spatio-tempo-ral network(GISTN)model to enhance the accuracy of traffic flow prediction.GISTN is the first model to apply graph isomorphism networks(GIN)to traffic flow prediction,innovatively integra-ting them with dual-scale temporal convolutional networks and gated recurrent units.This approach effectively captures complex nonlinear spatial dependencies and multiscale temporal patterns in traf-fic data.Experimental results on three public datasets demonstrate that GISTN consistently outper-forms classical baseline models across various prediction horizons.Overall,GISTN provides a novel and efficient solution for traffic flow prediction,offering significant implications for enhancing the performance and efficiency of intelligent transportation systems.关键词
交通流预测/图神经网络/图同构网络/时空建模/智能交通系统Key words
traffic flow prediction/graph neural networks/graph isomorphism network/spa-tio-temporal modeling/intelligent transportation systems分类
信息技术与安全科学引用本文复制引用
张伟阳,陈宏敏,林兵..一种基于图同构时空网络的交通流预测模型[J].福建师范大学学报（自然科学版）,2026,42(1):1-9,9.基金项目
国家自然科学基金项目(62072108) （62072108）
福建省高校产学合作项目(2022H6024、2021H6026) （2022H6024、2021H6026）
