Travel-mode inference based on GPS-trajectory data through multi-scale mixed attention mechanism

Area of Science:

• Computer Science
• Transportation Engineering
• Data Science

Background:

• Accurate travel mode identification is crucial for urban transportation planning.
• Global Positioning System (GPS) data offers significant potential for inferring travel modes.
• Existing methods may not fully capture the complex spatiotemporal dynamics of user movement.

Purpose of the Study:

• To develop an innovative method for inferring travel modes from GPS trajectory data.
• To leverage multi-scale convolutional techniques and attention mechanisms for enhanced accuracy.
• To improve the understanding of user movement and behavior patterns in urban environments.

Main Methods:

• Utilizing multi-scale convolutional neural networks to analyze spatiotemporal features in GPS data.
• Integrating an attention mechanism for autonomous learning and highlighting critical data points.
• Applying the method to the GeoLife open-source GPS trajectory dataset.

Main Results:

• The proposed method achieved an accuracy of 83.3% in travel mode inference.
• Demonstrated the effectiveness of combining multi-scale convolutions and attention mechanisms.
• Showcased improved prediction of user travel modes compared to baseline approaches.

Conclusions:

• The integrated approach of multi-scale convolutions and attention mechanisms significantly enhances travel mode inference accuracy.
• This method provides a robust framework for analyzing GPS trajectory data for transportation applications.
• The findings support the advancement of intelligent transportation systems through data-driven insights.