基于Transformer的短时交通流时空预测  

SHORT TERM TRAFFIC FLOW FORECASTING BASED ON TRASNSFORMER

作　　者：杨国亮 习浩 龚家仁 温钧林 Yang Guoliang;Xi Hao;Gong Jiaren;Wen Junlin(School of Electrical Engineering and Automation,Jiangxi University of Science and Technology,Ganzhou 341000,Jiangxi,China)

机构地区：[1]江西理工大学电气工程与自动化学院,江西赣州341000

出　　处：《计算机应用与软件》2024年第3期169-173,225,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(51365017);江西省教育厅科技计划项目(GJJ190450)。

摘　　要：现有的交通流预测模型未能全面获取路网的空间依赖,忽略了周期性对交通流量的影响,且缺乏对全局时间依赖的建模能力。针对以上问题,提出一种结合Transformer的动态扩散卷积门控循环单元预测模型。该模型利用动态扩散卷积网络和门控循环单元对交通流的近期、日周期和周周期三个时间进行时空建模;使用Transformer层获取全局时间依赖关系;将各组件输出进行加权融合,生成预测结果。实验结果表明,该方法相较基准模型能有效降低预测误差,准确预测交通演化态势。Existing traffic flow prediction models fail to fully obtain the spatial dependence of the road network,ignore the influence of periodicity to traffic data,and lack the ability to model global time dependence.To solve the above problems,a dynamic diffusion convolution and a gated recurrent unit prediction model combined with Transformer is proposed.The dynamic diffusion convolutional networks and gated recurrent unit were used to model the near-term,daily cycle and weekly cycle time of traffic flow.The Transformer layer was used to obtain the global time dependency.The output of each component was weighted and fused to generate the prediction result.The experimental results show that compared with the benchmark model,this method can effectively reduce the prediction error and accurately predict the traffic evolution situation.

关 键 词：短时交通流预测 扩散卷积 门控循环单元 TRANSFORMER 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]