ProDiff: Prototype-Guided Diffusion for Minimal Information Trajectory Imputation

Tianci Bu, Le Zhou, Wenchuan Yang, Jianhong Mou, Kang Yang, Suoyi Tan, Feng Yao, Jingyuan Wang, Xin Lu

TL;DR: A prototype-guided diffusion model that captures macro-level mobility patterns while maintaining high fidelity in minimal information trajectory imputation.

Abstract:

Trajectory data is crucial for various applications but often suffers from incompleteness due to device limitations and diverse collection scenarios. Existing imputation methods rely on sparse trajectory or travel information, such as velocity, to infer missing points. However, these approaches assume that sparse trajectories retain essential behavioral patterns, which place significant demands on data acquisition and overlook the potential of large-scale human trajectory embeddings.
To address this, we propose ProDiff, a trajectory imputation framework that uses only two endpoints as minimal information. It integrates prototype learning to embed human movement patterns and a denoising diffusion probabilistic model for robust spatiotemporal reconstruction. Joint training with a tailored loss function ensures effective imputation.
ProDiff outperforms state-of-the-art methods, improving accuracy by 6.28\% on FourSquare and 2.52\% on WuXi. Further analysis shows a 0.927 correlation between generated and real trajectories, demonstrating the effectiveness of our approach.

Lay Summary:

Think about GPS tracks from phones or cars; they often have missing parts. This makes it hard to see the full picture of how people move around, which is important for things like planning city services. Current ways to fill in these gaps often need a lot of the original path to be there.

We have created a smart system called ProDiff that can draw a likely path even if it only knows where a journey started and ended. It works by first learning the common shapes and patterns of how people usually travel. Then, it uses this knowledge to make an educated guess to fill in the blanks.
ProDiff is much better at guessing the missing parts of a journey compared to older methods. This means we can get more complete and accurate information about movement, even from spotty data. This can lead to better traffic management, improved public transport planning, and a clearer understanding of how people use a city.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/b010001y/ProDiff

Primary Area: Applications->Time Series

Keywords: Minimal Information Trajectory Imputation, Prototype Learning, Diffusion Model

Flagged For Ethics Review: true

Submission Number: 8921