基于LSTM模型的在线地图路段交通状态预测  

Road Traffic Status Prediction of Online Map Based on LSTM Model

作　　者：胡钦 王庆国 HU Qin;WANG Qingguo(School of Automotive and Traffic Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]武汉科技大学汽车与交通工程学院,湖北武汉430081

出　　处：《物流科技》2023年第7期62-66,106,共6页Logistics Sci-Tech

基　　金：国家自然科学基金项目(41571396)。

摘　　要：为准确及时地预测城市道路交通状态,帮助管理部门实施交通管理措施,预防交通拥堵发生,文章实时获取在线地图交通状态数据,将其划分为路段粒度后,使用路段上下游以及对向车道交通状态作为特征矩阵输入LSTM网络模型,对路段工作日的交通状态进行预测,并与单路段交通状态作为特征矩阵输入的结果和其他模型结果做对比。实验结果表明,考虑多路段LSTM网络模型预测的平均MAE、RMSE和准确率分别为3.797、6.263和82.15%,证明了LSTM网络模型能较好地预测对路段状态,且考虑到路段上下游车道状态因素相对于单纯考虑路段的交通状态可以提高预测精度。In order to accurately and timely predict the state of urban road traffic,help management sectors to implement traffic management measures to prevent traffic congestion.In this paper,the traffic state of online map are obtained in real time,and they are divided into section granularity.Then,the traffic state of upstream and downstream sections and the traffic state of opposite lanes are used as the characteristic matrix to input the LSTM network model,and the traffic state of section working days is predicted.The results are compared with the results of single section traffic state as the feature matrix input and other model results.The experimental results show that the average MAE,RMSE and accuracy of the LSTM model are 3.797,6.263 and82.15%,respectively,which proves that the LSTM model can better predict traffic state of the road section,and the prediction accuracy can be improved by considering the traffic state of the upstream and downstream lanes of the road section compared with the traffic state of the road section.

关 键 词：交通状态预测 深度学习 LSTM模型 数据挖掘 城市道路 

分 类 号：U491.4[交通运输工程—交通运输规划与管理]