摘要
实时、准确的交通流短期预测是交通诱导、管理的前提.为了提高预测精度,结合交通流数据中的历史时间相关性与网络空间断面相关性,构建了一种基于皮尔森相关系数法(Pearson Correlation Coefficient,PCC)与双向长短时记忆(Bidirectional Long Short Term Memory,BLSTM)架构的交通流短时预测模型.该模型可以通过PCC筛选路网中与目标路段空间相关的路段,并将其重构为新数据集,作为BLSTM预测模型的输入,以实现交通流短期预测.通过美国加州交通流数据对模型预测性能进行评价,实验结果表明:该模型可以融合交通流数据中的时空相关性,相对于其他主流预测模型精度平均可提高4.83%.
Real-time and accurate short-term traffic flow prediction is the premise of traffic guidance and management.In order to improve the prediction accuracy,a short-term traffic flow prediction model based on Pearson Correlation Coefficient(PCC)method and Bidirectional Long Short Term Memory(BLSTM)architecture is constructed by combining the historical temporal correlation of traffic flow data with the cross-sectional correlation of network space.The model can use the PCC method to select the road section in the road network that is spatially related to the target road section,and reconstruct it into a new data set as the input of the BLSTM prediction model to achieve short-term traffic flow prediction.The performances of the proposed models are validated by using the traffic flow data that are obtained from Caltrans Performance Measurement System(PeMS)in the United States.The research results show that the model can integrate the temporal and spatial correlations in traffic flow data.Compared with other mainstream prediction models,the prediction accuracy of the proposed model can be improved by at least 4.83%.
作者
邵春福
薛松
董春娇
王晟由
庄焱
SHAO Chunfu;XUE Song;DONG Chunjiao;WANG Shengyou;ZHUANG Yan(Key Laboratory of Transport Industry of Big Data Application Technologies for Comprehensive Transport,Beijing Jiaotong University,Beijing 100044,China)
出处
《北京交通大学学报》
CAS
CSCD
北大核心
2021年第4期37-43,共7页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
基金
国家自然科学基金(71621001)。
关键词
智能交通
交通流短期预测
神经网络
相关性分析
intelligent transportation
short-term traffic flow prediction
neural network
correlation analysis
