According to the traffic flow prediction method based on the space-time convolution fusion probability sparse attention mechanism, a space-time diagram convolution module (T-GCN Block), a space-time convolution module (ST-Conv Block) and a multi-head probability sparse attention module (ProbSSATt Block) are included in the traffic flow combined prediction method. Wherein the Gated TCN and the GCN in the T-GCN Blocks are respectively used for capturing the time dependency and the spatial correlation of the traffic flow, and a plurality of T-GCN Blocks are stacked to process the spatial dependencies of different time levels; sT-Conv Block is used for capturing complex time dependence of traffic flows at the same position and dynamic spatial correlation of traffic flows at adjacent positions on the same time step length; and the ProbSSATt Block is combined with the dynamic spatio-temporal characteristics, and long-term prediction is effectively carried out.
一种时空卷积融合概率稀疏注意力机制的交通流预测方法,所述的交通流组合预测方法,包含时空图卷积模块(T‑GCN Block)、时空卷积模块(ST‑Conv Block)和多头概率稀疏注意力模块(ProbSSAtt Block)。其中,T‑GCN Block中的Gated TCN和GCN分别用于捕捉交通流的时间依赖性和空间相关性,堆叠多个T‑GCN Block以处理不同时间级别的空间依赖性;ST‑Conv Block用于捕获同一位置交通流的复杂时间依赖性和同一时间步长上邻近位置交通流的动态空间相关性;ProbSSAtt Block结合动态时空特征并有效地进行长期预测。
Traffic flow prediction method of space-time convolution fusion probability sparse attention mechanism
一种时空卷积融合概率稀疏注意力机制的交通流预测方法
2023-12-26
Patent
Elektronische Ressource
Chinesisch
IPC: | G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS |
Europäisches Patentamt | 2023
|Europäisches Patentamt | 2023
|Europäisches Patentamt | 2023
|Attention-based dynamic space-time diagram convolution traffic flow prediction system and method
Europäisches Patentamt | 2023
|Transportation Research Record | 2024
|