School of Computer Science and Engineering, Hunan University of Science and Technology
Hunan Provincial Key Laboratory for Service Computing and Novel Software Technology,Hunan University of Science and Technology
Changsha Institute of Technology

及时、准确的交通流预测对于车辆导航规划和智能交通调度具有重要作用.道路交通不仅具有时空相关性,同时多种环境因素还会对交通状况造成重要影响.为提高道路交通流速度预测的准确率,首先对试验选择地区的降雨程度、空气污染程度数据进行分级处理,时间周期划分工作日与非工作日;然后结合双向长短期记忆网络和特征工程技术,建立基于多影响因素的交通流速度预测模型(MF-TPM),并在公开的地区交通速度时序数据集与天气数据集上建模分析;最后基于大规模真实交通数据进行试验,结果表明:MF-TPM的预测精度比常用的长短期记忆网络、卷积神经网络和双向长短期记忆网络模型分别高2.20%,4.94%和0.63%.在不同的降雨程度和空气污染等级下,MF-TPM同样具有最佳的预测表现.

Timely and accurate traffic flow prediction plays an important role in navigation planning andintelligent traffic dispatch. Road traffic is not only temporal and spatial correlation,but also a variety ofenvironmental factors will have an important impact on traffic conditions. In order to improve the accuracy of roadtraffic flow speed prediction, firstly, the data of rainfall degrees and air pollution levels are graded, and the timeperiod is divided into working days and non-working days. Then combined with Bidirectional Long-Term andShort-Term Memory network ( BiLSTM) and feature engineering technology, we need to establish a Multi-Factor-based Traffic Flow Speed Prediction Model ( MF-TPM) , and model and analyze the public regional traffic speedtime series datasets and weather datasets. Finally, experiments based on large-scale real traffic data show that theprediction accuracy of MF-TPM is 2.20%, 4.94% and 0.63% higher than the commonly used Long-Short-TermMemory network ( LSTM) , Convolutional Neural Network ( CNN) and BiLSTM network models, respectively.MF-TPM also has the best prediction performance under different rainfall levels and air pollution levels.

湖南省自然科学基金资助项目(2021JJ30276);湖南省教育厅科学研究重点项目资助(19A172);国家自然科学基金资助项目(41871320)

刘思林,廖祝华,符琦,等.多影响因素下的交通流速度预测[J].湖南科技大学学报(自然科学版),2024,39(3):109-115.

LIU S L, LIAO Z H, FU Q, et al. Prediction of Traffic Flow Speed Under Multiple Influencing Factors [ J] . Journal of HunanUniversity of Science and Technology ( Natural Science Edition) , 2024, 39( 3) : 109-115.