一种基于图自注意力机制与霍克斯过程的交通流预测方法
By adopting a traffic flow prediction method based on graph self-attention mechanism and Hawkes process, the problem that existing models cannot effectively capture the dynamic spatiotemporal dependence of traffic flow is solved, and higher accuracy traffic flow prediction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG NORMAL UNIV
- Filing Date
- 2022-09-14
- Publication Date
- 2026-07-17
AI Technical Summary
Existing time series models and Kalman filters cannot effectively capture long-term dependencies. Existing technologies cannot effectively address the highly nonlinear characteristics and dynamic time series models of traffic flow, nor can they effectively capture the dynamic spatiotemporal and long-term dependencies of traffic flow, resulting in low traffic prediction accuracy.
A traffic flow prediction method based on graph self-attention mechanism and Hawkes process is adopted. By acquiring the structural embedding information of traffic flow graph, spatial features are captured by graph convolutional neural network and combined with temporal and category embedding features. Attention score is calculated through self-attention mechanism, approximating the conditional strength function of Hawkes process. Monte Carlo sampling is used to predict the occurrence time and category of future events.
It improves the accuracy of traffic flow forecasting, better captures long-sequence dependencies and spatial information, and enhances the predictive performance of future events.
Smart Images
Figure CN115358485B_ABST