The invention relates to the technical field of traffic management systems, and discloses a traffic flow prediction method and system based on digraph spatio-temporal information embedding, and the method comprises the steps: constructing a digraph based on a road network structure of a one-way traffic flow, and enhancing the spatial feature extraction capability of a model for traffic spatio-temporal data; static space information and dynamic space information based on a directed graph are embedded into traffic flow data, and the learning ability of the model for road network space information is enhanced; a spatio-temporal attention memory block with time period information embedding is designed to capture spatio-temporal correlation of data. In the encoder stage, traffic spatio-temporal data features are extracted according to the spatio-temporal information embedding method and the spatio-temporal attention memory block connection design; in the decoder stage, a final prediction result is generated by using a full convolution UniRepLKNet-S architecture. The method has the advantages that traffic management departments and individual drivers can be better helped to cope with complex and changeable traffic environments, and the road network traffic efficiency is improved.
本发明涉及交通管理系统技术领域,公开了一种基于有向图时空信息嵌入的交通流预测方法及系统,基于单向交通流的路网结构,构造有向图,加强模型对交通时空数据的空间特征提取能力;将基于有向图的静态空间信息和动态空间信息嵌入到交通流数据中,增强模型对路网空间信息的学习能力;设计了具有时间周期信息嵌入的时空注意力记忆块来捕捉数据的时空相关性。编码器阶段,根据上述的时空信息嵌入方法和时空注意力记忆块连接设计以提取交通时空数据特征;解码器阶段,使用全卷积的UniRepLKNet‑S架构生成最终的预测结果。本发明的有益效果为:能够更好的帮助交通管理部门和个体驾驶者应对复杂多变的交通环境,提高路网交通效率。
Traffic flow prediction method and system based on directed graph spatio-temporal information embedding
一种基于有向图时空信息嵌入的交通流预测方法及系统
2024-07-05
Patent
Electronic Resource
Chinese
Traffic flow prediction method for traffic flow spatio-temporal data information
European Patent Office | 2024
|Traffic Flow Prediction Based on Spatio-Temporal Aggregated Graph Neural Networks
Transportation Research Record | 2025
|Spatio-temporal graph attention networks for traffic prediction
Taylor & Francis Verlag | 2024
|