TraceBERT-A Feasibility Study on Reconstructing Spatial-Temporal Gaps from Incomplete Motion Trajectories via BERT...

Area of Science:

• Mobility Data Science
• Geographic Information Science
• Artificial Intelligence

Background:

• Trajectory data are crucial for understanding human mobility and travel behaviors.
• Data collection issues often lead to spatial-temporal gaps and missing trajectory segments.
• Reconstructing non-repetitive individual motion traces is challenging due to the lack of historical data.

Purpose of the Study:

• To address the problem of missing trajectory data by proposing a novel framework.
• To adapt advanced natural language processing techniques for trajectory reconstruction.
• To evaluate the effectiveness of the proposed method on real-world mobility data.

Main Methods:

• Leveraging Bidirectional Encoder Representations from Transformers (BERT), a state-of-the-art language representation model.
• Training deep bidirectional representations from unlabeled location sequences, conditioned on both left and right context.
• Developing a framework named TraceBERT for trajectory processing and reconstruction.

Main Results:

• TraceBERT demonstrated effective trajectory reconstruction by predicting missing locations.
• The framework was tested on a large-scale real-world dataset of short-term tourist trajectories.
• The proposed method showed prominent potential compared to traditional statistical approaches.

Conclusions:

• Advanced language modeling approaches can be successfully adapted for mobility-based applications.
• TraceBERT offers a promising solution for filling spatial-temporal gaps in trajectory data.
• The study highlights the potential of AI in enhancing the analysis of human mobility patterns.