GLSNN Network: A Multi-Scale Spatiotemporal Prediction Model for Urban Traffic Flow

Area of Science:

• Intelligent Transportation Systems
• Machine Learning
• Data Science

Background:

• Traffic flow prediction is crucial for intelligent transportation systems.
• Big data presents challenges and opportunities for traffic prediction models.
• Urban traffic speed prediction is a complex spatiotemporal problem.

Purpose of the Study:

• To develop a novel model for multi-scale spatiotemporal fusion prediction of urban traffic flow.
• To address the challenges of big data in traffic prediction.
• To improve the accuracy and scale of traffic speed predictions.

Main Methods:

• A Graph Long Short-Term Memory (LSTM) Spatiotemporal Neural Network (GLSNN) was constructed.
• The GLSNN model integrates multi-source input data.
• Key components include MS-LSTM for temporal scaling, and LZ-GCN and LSTM-GRU for spatiotemporal dependency capture.

Main Results:

• The GLSNN model demonstrated superior performance in urban traffic flow prediction.
• The model achieved high-precision predictions.
• The model successfully performed multi-scale spatiotemporal fusion prediction.

Conclusions:

• The GLSNN model offers a robust solution for urban traffic speed prediction in the big data era.
• The proposed model effectively captures complex spatiotemporal dependencies.
• Experimental results validate the model's effectiveness and superior performance over existing methods.