多影响因素下的交通流速度预测

Prediction of Traffic Flow Speed Under Multiple Influencing Factors

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 and intelligent traffic dispatch.Road traffic is not only temporal and spatial correlation,but also a variety of environmental factors will have an important impact on traffic conditions.In order to improve the accuracy of road traffic flow speed prediction,firstly,the data of rainfall degrees and air pollution levels are graded,and the time period is divided into working days and non-working days.Then combined with Bidirectional Long-Term and Short-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 speed time series datasets and weather datasets.Finally,experiments based on large-scale real traffic data show that the prediction accuracy of MF-TPM is 2.20%,4.94%and 0.63%higher than the commonly used Long-Short-Term Memory 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.