IPGMVL: based on interactive progressive graph convolution with multi-view learning traffic flow forecasting.

Area of Science:

• Artificial Intelligence
• Transportation Engineering
• Data Science

Background:

• Traffic flow prediction is crucial for intelligent transportation systems.
• Existing graph neural networks struggle with dynamic network changes and spatial heterogeneity.
• Accurate modeling of complex spatio-temporal dependencies remains a key challenge.

Purpose of the Study:

• To propose a novel model, Interactive Progressive Graph Convolution with Multi-view learning (IPGMVL), for enhanced traffic flow prediction.
• To address limitations of static graph construction and neglected interactive dependencies in current models.
• To improve the adaptability and accuracy of traffic prediction systems.

Main Methods:

• Developed a progressive graph convolution for dynamic edge weight adjustment based on trend similarity.
• Implemented a multi-view interactive learning mechanism to model spatio-temporal heterogeneous patterns.
• Introduced fast parallel learning (FPL) and serial learning (SL) modules for efficient spatio-temporal feature extraction and enhanced dependency modeling.

Main Results:

• IPGMVL achieved superior performance on four benchmark datasets, setting new state-of-the-art standards.
• The model demonstrated effective capture of real-time spatial evolution and complex spatio-temporal dependencies.
• IPGMVL maintained computational efficiency while improving prediction accuracy.

Conclusions:

• Dynamic graph adaptation and interactive learning are vital for advanced traffic prediction.
• The proposed IPGMVL model offers a significant advancement in modeling spatio-temporal dependencies for traffic flow.
• This research provides a more robust and efficient solution for real-world traffic prediction applications.