基于时空相关性和遗传小波神经网络的路网短时交通流预测  

Short-term Traffic Flow Forecasting of Road Network Based on Spatiotemporal Correlation and Genetic Wavelet Neural Network

作　　者：陆百川[1,2] 李玉莲 舒芹 LU Baichuan;LI Yulian;SHU Qin(School of Traffic&Transportation,Chongqing Jiaotong University,Chongqing 400074,China;Key Lab of Traffic System&Safety in Mountain Cities,Chongqing 400074,China)

机构地区：[1]重庆交通大学交通运输学院,重庆400074 [2]重庆山地城市交通系统与安全实验室,重庆400074

出　　处：《重庆理工大学学报（自然科学）》2020年第5期25-34,共10页Journal of Chongqing University of Technology：Natural Science

基　　金：中国博士后科学基金面上项目(2016M592645)。

摘　　要：根据城市路段交通流在时间维度的变化规律和在空间维度的分布特征,以及智能算法对交通流数据的较强适应性和降噪能力,提出了基于时空相关性和遗传小波神经网络(GA-WNN)的路网短时交通流预测。首先,分析了路网交通流的时空特性和数据特征,建立了适用于交通路网的空间邻接矩阵;其次,以时空相关性函数量化不同时间延迟下路段与周边相邻路段交通状态之间的影响程度,并将其作为交通流预测模型输入变量的判断指标,结合遗传算法的全局搜索及小波神经网络的自适应学习优点构建了交通流预测模型;最后,通过实例验证表明,基于GA-WNN的交通流预测方法比其他方法更有优势,对比单一时间序列和空间序列预测方法,考虑了交通流时空相关性的预测方法能提高短时交通流预测精度。According to the change rule of traffic flow in time dimension and its distribution characteristics in space dimension,and the strong adaptability and noise reduction ability of intelligent algorithm to traffic flow data,this paper proposes a short-term traffic flow prediction based on spatialtemporal correlation and genetic wavelet neural network( GA-WNN). Firstly,the temporal and spatial characteristics and data characteristics of traffic flow in road network are analyzed,and the spatial adjacency matrix suitable for traffic network is established. Secondly,the spatial and temporal correlation function is used to quantify the degree of influence between the traffic state of road segment and adjacent road segment with different time delays,and it is used as the judgment index of the input variable of traffic flow prediction model,and a traffic flow prediction model based on adaptive learning of wavelet neural network is constructed,combined with the global search of genetic algorithm and the global search of genetic algorithm. Finally,an example is given to show that the traffic flow prediction method based on GA-WNN has more advantages than other methods. At the same time,compared with single time series and spatial series prediction methods,the prediction method considering the temporal-spatial correlation of traffic flow can effectively improve the accuracy of short-term traffic flow prediction.

关 键 词：短时交通流预测 时空相关性 交通流数据 空间邻接矩阵 遗传小波神经网络 

分 类 号：U491.14[交通运输工程—交通运输规划与管理]