The invention relates to the technical field of intelligent traffic, in particular to a highway service area traffic flow prediction method, a time-space diagram convolutional neural network model based on an attention mechanism belongs to an STGNN model based on CNNs, ChebNet diagram convolution and one-dimensional convolution along the time dimension are creatively combined, and the prediction problem of time-space traffic diagram data is solved. A structure multi-head self-attention module is designed, and road network topological structure information is extracted; designing a dynamic evolution graph convolution module, learning a new graph for each time slice, and adaptively and dynamically adjusting the correlation strength between nodes by using a self-attention mechanism; and designing a time multi-head self-attention module and time position embedding, and extracting time correlation. And a Transform structure is optimized, so that the Transs-STGNN has a more accurate long-term prediction capability. According to the method, the space-time correlation can be better extracted, and the prediction precision can be improved.
本发明涉及智能交通技术领域,尤其涉及一种高速公路服务区车流量预测方法,基于注意力机制的时空图卷积神经网络模型属于基于CNNs的STGNN模型,创造性地将ChebNet图卷卷积和沿时间维度的一维卷积相结合,用于解决时空交通图数据的预测问题。设计结构多头自注意模块,提取路网拓扑结构信息;设计动态演化图卷积模块,为每个时间片学习一个新的图,并利用自注意机制自适应地动态调整节点之间的相关强度;设计时间多头自注意模块和时间位置嵌入,提取时间相关性。并通过优化Transformer结构,使得Trans‑STGNN有更准确的长期预测能力。本方法能够更好地提取时空相关性,并能提高预测精度。
Expressway service area traffic flow prediction method
一种高速公路服务区车流量预测方法
2023-08-18
Patent
Electronic Resource
Chinese
Expressway intelligent service area traffic flow prediction method and management method
European Patent Office | 2024
|Expressway interleaving area traffic accident prediction method
European Patent Office | 2021
|Expressway service area traffic guidance method and system
European Patent Office | 2021
|