Spatio-temporal information enhance graph convolutional networks: A deep learning framework for ride-hailing demand...

Area of Science:

• Urban planning and transportation science.
• Artificial intelligence and machine learning.
• Geospatial data analysis.

Background:

• Accurate ride-hailing demand prediction is vital for optimizing urban transportation systems, including vehicle scheduling and traffic management.
• Existing methods struggle to effectively integrate external spatiotemporal factors, leading to prediction inaccuracies.
• The influence of external factors like weather and events on ride-hailing demand is significant but often underestimated.

Purpose of the Study:

• To develop an advanced demand prediction model that overcomes the limitations of previous approaches by effectively incorporating external spatiotemporal influences.
• To enhance the accuracy and timeliness of ride-hailing demand forecasts.
• To improve the model's ability to capture complex spatiotemporal dependencies.

Main Methods:

• Proposed a novel spatiotemporal information-enhanced graph convolution network (STE-GCN) for ride-hailing demand prediction.
• Utilized correlation analysis to extract and encode key external spatiotemporal factors into area-specific feature units.
• Employed gated recurrent units (GRUs) and graph convolutional networks (GCNs) to model spatiotemporal dependencies between demand and external factors.

Main Results:

• The proposed STE-GCN model demonstrated superior prediction performance compared to baseline models when external spatiotemporal factors were integrated.
• The model exhibited consistent accuracy across different experimental areas, indicating robustness.
• Incorporating external factors significantly enhanced the model's prediction accuracy and its perceptiveness to real-world influences.

Conclusions:

• External spatiotemporal factors are crucial for improving the performance of ride-hailing demand prediction models.
• The developed STE-GCN model is robust and offers excellent performance, highlighting its broad application potential in intelligent transportation systems.
• This research provides a valuable tool for optimizing ride-hailing services and mitigating urban traffic congestion.