燕山大学学报2025，Vol.49Issue(2)：95-105,11.DOI:10.3969/j.issn.1007-791X.2025.02.001
基于时空图神经网络的城市路网行程时间预测研究综述
Review on travel time prediction of urban road network based on spatial-temporal graph neural network
摘要
Abstract
With urbanization accelerating and traffic networks continuously expanding,urban traffic management faces increasingly complex challenges.Accurate travel time prediction is essential for optimizing traffic management,enhancing travel experiences,and advancing smart city development.The travel time of urban road networks exhibits pronounced spatio-temporal dependence and randomness due to the complex network structure,dynamic traffic flow changes,and external factors.Spatio-temporal graph neural networks serve as powerful tools for modeling such complexities,effectively capturing intricate spatial-temporal relationships in urban road networks.Consequently,constructing a travel time prediction framework based on spatio-temporal graph neural networks has emerged as a key research focus in smart transportation.Starting from the key elements of the travel time prediction framework based on spatio-temporal graph neural network,specifically,spatiotemporal information modeling,prediction task selection,and learning paradigm design,the research progress of this type of research in the past three years is introduced.Firstly,the research in this area is summarized from the definition of the problem and the basic framework.Subsequently,based on the different number of prediction task selections in key elements,the related research is categorized into two groups:single-task and multi-task travel time prediction methods.This division is used to explore in detail the unique characteristics and representative works associated with each type of prediction method.Finally,the difficulties in modeling spatio-temporal high-order correlations,implicit spatio-temporal dependencies,and interpretability aspects of travel time prediction are discussed,along with future development trends.关键词
图神经网络/时空图序列/时空数据挖掘/行程时间预测Key words
graph neural network/spatio-temporal graph sequence/spatial-temporal data mining/travel time prediction分类
计算机与自动化引用本文复制引用
董慧,潘晓,郭景峰,陈晓,王书海..基于时空图神经网络的城市路网行程时间预测研究综述[J].燕山大学学报,2025,49(2):95-105,11.基金项目
河北省自然科学基金资助项目(F2021210005,F2023407003,F2024210042) （F2021210005,F2023407003,F2024210042）
河北省研究生创新基金项目(CXZZBS2022117) （CXZZBS2022117）
河北省海洋动力过程与资源环境重点实验室开放课题(HBHY02) （HBHY02）
