Spatial-temporal hypergraph convolutional network for traffic forecasting.

Area of Science:

• Intelligent Transportation Systems
• Data Science
• Network Analysis

Background:

• Traffic forecasting is crucial for intelligent transportation systems.
• Existing methods struggle with spatial isomorphism and temporal drift in traffic data.
• Accurate forecasting requires capturing intricate spatial-temporal dependencies.

Purpose of the Study:

• To propose a novel spatial-temporal hypergraph convolutional network (ST-HCN) for enhanced traffic forecasting.
• To address limitations in current methods regarding road-network isomorphism and periodic temporal drift.
• To improve the accuracy and efficiency of traffic prediction models.

Main Methods:

• Developed a spatial-temporal hypergraph convolutional network (ST-HCN).
• Employed K-means clustering and physical road network characteristics to unify spatial correlations.
• Utilized a dual-channel hypergraph convolution for high-order spatial relationships.
• Incorporated a Convolutional Long Short-Term Memory (ConvLSTM) network to handle temporal drift.

Main Results:

• The proposed ST-HCN framework effectively captures high-order spatial relationships.
• The model successfully addresses the periodic drift problem in traffic data.
• Experimental results demonstrate superior performance compared to state-of-the-art baseline methods.
• Validated the framework's efficacy on real-world traffic datasets.

Conclusions:

• The ST-HCN model offers a significant advancement in traffic forecasting accuracy.
• The approach effectively models complex spatial-temporal dependencies in traffic networks.
• This work provides a robust solution for intelligent transportation systems.