Spatiotemporal-decoupled interactive learning for traffic flow prediction

Area of Science:

• Intelligent Transportation Systems
• Data Science
• Network Science

Background:

• Accurate traffic flow prediction is crucial for intelligent transportation systems (ITS).
• Existing methods often fail to capture complex spatiotemporal dependencies and diverse patterns due to spatial heterogeneity and temporal variations.
• This limits the effectiveness of trip planning, network dispatch, and management decisions.

Purpose of the Study:

• To propose a novel framework, Spatiotemporal-Decoupled Interactive Learning (STDIL), to address the limitations of existing traffic flow prediction methods.
• To enhance the learning of spatiotemporal dependencies and accommodate pattern diversity in traffic flow data.
• To improve the accuracy and adaptability of traffic flow prediction models.

Main Methods:

• The proposed STDIL framework integrates a spatiotemporal decoupling module and an interactive learning module.
• The spatiotemporal decoupling module reconstructs sequences along spatial and temporal dimensions for discriminative representations.
• The interactive learning module dynamically reconstructs graph structures to capture global and local spatiotemporal correlations, including long-range dependencies.

Main Results:

• Experiments on four real-world urban traffic flow datasets demonstrated that STDIL significantly outperforms existing methods across all prediction horizons.
• STDIL effectively handles spatiotemporal heterogeneity and dynamic dependencies inherent in traffic data.
• The framework shows adaptability to diverse traffic scenarios, achieving higher prediction accuracy.

Conclusions:

• STDIL provides a more effective approach to traffic flow prediction by addressing limitations in existing methods.
• The framework's ability to capture complex spatiotemporal interactions leads to significant accuracy improvements.
• STDIL offers a promising solution for enhancing the capabilities of intelligent transportation systems.