联合时空特征的交通流参数预测综述  

A review on parameters prediction of traffic flow by combining spatio-temporal features

作　　者：关为生 肖建力[1] GUAN Weisheng;XIAO Jianli(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《上海理工大学学报》2022年第6期592-602,共11页Journal of University of Shanghai For Science and Technology

基　　金：国家自然科学基金资助项目(61603257,61906121)。

摘　　要：针对交通流参数预测在智能交通系统中的重要性,为寻求更实时准确的预测方法,对联合时空特征的交通流参数预测方法进行综述。以交通时空数据为研究对象,将交通流参数预测方法归纳为统计学习方法、深度学习方法和图神经网络方法。基于这3类方法分别从传统和联合时空特征角度概括了各种方法的研究现状和特点,分析了交通流参数预测的难点。结果表明,联合时空特征的交通流预测方法由于考虑了道路网络中复杂且动态的时空依赖性,相较于传统的同类方法,预测性能有较大提升。最后,从模型输入和模型设计角度,讨论了交通流参数预测未来研究方向。Considering the importance of traffic flow parameters prediction in intelligent transportation systems,the traffic flow parameter prediction method combining spatio-temporal features was summarized to find a more accurate real-time prediction method.Taking the traffic spatio-temporal data as the research object,the prediction methods of traffic flow parameters were divided into statistical learning method,deep learning method,and graph neural network method.Based on these three categories methods,the research status and characteristics of various methods were summarized.The difficulties of traffic flow parameters prediction were analyzed from the perspective of traditional and spatio-temporal features.The results show that the traffic flow prediction method combined with spatiotemporal features has a great improvement in prediction performance compared with the traditional similar methods because it considers the complex and dynamic spatio-temporal dependence in the road network.Finally,the future research direction of traffic flow parameters prediction was discussed from the perspective of model input and model design.

关 键 词：交通流预测 统计学习 深度学习 图神经网络 

分 类 号：U491.112[交通运输工程—交通运输规划与管理]