基于多组件融合与空洞图卷积的车道占用率预测模型

为解决交通拥堵和交通硬件资源分配不足等问题，提出一种基于多组件融合与空洞图卷积的车道占用率预测模型MCFDGCN。针对交通数据的非线性和受多种隐式因素影响的特点，利用图卷积提取交通数据的空间相关性，使用空洞卷积提取时间依赖特征，将车流量和车辆速度作为2个隐式因素引入模型中，对多组件提取的影响车道占用率的多模态特征进行融合，以完成车道占用率预测任务。在PeMS7（O）、PeMS7（4）数据集上进行实验，结果表明，与HA、ARIMA等模型相比，MCFDGCN模型预测误差较低且误差增长较缓慢，能实现更精准的车道占用率预测。

Lane Occupancy Prediction Model Based on Multi-Component Fusion and Dilated Graph Convolution

In order to solve the problems of traffic congestion and insufficient allocation of traffic hardware resources,a lane occupancy prediction model,MCFDGCN,based on multi-component fusion and dilated graph convolution is proposed.Considering the non-linear feature of traffic data,and that it is influenced by many implicit factors,the model uses graph convolution to extract the spatial correlation of traffic data,and uses dilated convolution to extract time-dependent features.The vehicle flow and vehicle speed are added into the model as two implicit factors,and the multimodal features extracted by multiple components are fused to predict the lane occupancy rate.Experimental results on the datasets of PeMS7(O) and PeMS7(4) show that compared with HA,ARIMA and other models,the MCFDGCN model exhibits a lower average error rate and error growth rate.It can achieve more accurate prediction of the lane occupancy rate.