摘要
交通速度是衡量交通状态的一个重要指标,实时、准确的交通速度预测是构建智能交通系统的重要一环。针对交通速度存在随机性、非线性、时空相关性等问题,提出了一种新的基于注意力机制和图卷积神经网络相结合的交通速度预测模型。首先,使用注意力机制构建时空注意力权重矩阵,再联合图卷积方法捕获交通信息中的空间相关性特征;然后,通过门控时卷积的方法获取时间相关性;最后,将所提模型与其他5个基准模型分别在2组公开的交通速度数据集上进行预测。实验结果表明,该预测模型在2个数据集上的准确率分别为75.1%和86.6%,比先进的基准模型的准确率高3%左右。说明所提模型具有较高的准确率和稳定性,可为交通管理提供科学依据。
Traffic speed is an important indicator to measure the traffic state.Real-time and accurate traffic speed prediction is an important part of building an intelligent traffic system.Towards the problems of the randomness,nonlinearity and spatio-temporal correlation of traffic speed,a traffic speed prediction model is proposed by using the attention based spatio-temporal graph convolutional neural network.Firstly,the spatial and temporal attention weight matrix is constructed by using the attention mechanism,with the spatial correlation features of traffic information captured by combining with the graph convolution method.Secondly,the time correlation is obtained by gating time convolution.Finally,the proposed model and the other five benchmark models are tested on two public traffic speed data sets respectively.Experimental results show that the accuracy of the prediction model is 75.1%and 86.6%on the two data sets respectively,which is about 3%higher than the advanced benchmark models.It shows that the proposed model can provide important scientific basis for traffic control with its high accuracy and stability.
作者
黄伟坚
李春贵
HUANG Weijian;LI Chungui(School of Electrical,Electronic and Computer Science,Guangxi University of Science and Technology,Liuzhou 545616,China)
出处
《广西科技大学学报》
2022年第1期54-62,共9页
Journal of Guangxi University of Science and Technology
基金
广西自然科学基金项目(2018GXNSFAA050020)资助。
关键词
交通速度预测
注意力机制
图卷积
时空相关性
traffic speed prediction
attention mechanism
graph convolution
spatial-temporal correlation
