An Efficient Short-Term Traffic Speed Prediction Model Based on Improved TCN and GCN

Area of Science:

• Intelligent Transportation Systems (ITS)
• Traffic flow dynamics
• Machine learning for spatiotemporal prediction

Background:

• Accurate traffic speed prediction is crucial for effective traffic control and guidance within Intelligent Transportation Systems (ITS).
• Traffic speed is influenced by complex temporal and spatial dependencies, road network topology, and external factors.
• Existing models often struggle to capture the intricate spatiotemporal dynamics and network structure effectively.

Purpose of the Study:

• To propose a novel Speed Prediction of Traffic Model Network (SPTMN) for enhanced short-term traffic speed forecasting.
• To integrate advanced deep learning techniques for capturing complex traffic patterns.
• To improve the accuracy and reliability of traffic speed predictions.

Main Methods:

• The proposed SPTMN model leverages a Temporal Convolutional Network (TCN) for extracting temporal and local spatial features.
• A Graph Convolutional Network (GCN) is employed to capture the global spatial dependencies and topological relationships within the road network.
• Combined spatiotemporal features are integrated with road parameters for prediction.

Main Results:

• The SPTMN model demonstrated superior performance across diverse road conditions compared to eight baseline methods.
• Prediction errors were reduced by a minimum of 8% using the SPTMN model.
• The model exhibited high effectiveness and stability in real-world traffic scenarios.

Conclusions:

• The SPTMN model effectively integrates temporal and spatial features for accurate short-term traffic speed prediction.
• The hybrid TCN-GCN architecture provides a robust solution for the complexities of traffic flow dynamics.
• SPTMN offers significant improvements in prediction accuracy and reliability for Intelligent Transportation Systems.