Towards Robust Trajectory Representations: Isolating Environmental Confounders with Causal Learning
This work addresses the issue of limited generalization in trajectory modeling for mobility pattern analysis, representing an incremental improvement by applying causal methods to a known bottleneck.
The paper tackled the problem of geospatial context confounding trajectory modeling by proposing a causal learning framework (TrajCL) that isolates environmental confounders, resulting in enhanced performance in trajectory classification tasks with superior generalization and interpretability.
Trajectory modeling refers to characterizing human movement behavior, serving as a pivotal step in understanding mobility patterns. Nevertheless, existing studies typically ignore the confounding effects of geospatial context, leading to the acquisition of spurious correlations and limited generalization capabilities. To bridge this gap, we initially formulate a Structural Causal Model (SCM) to decipher the trajectory representation learning process from a causal perspective. Building upon the SCM, we further present a Trajectory modeling framework (TrajCL) based on Causal Learning, which leverages the backdoor adjustment theory as an intervention tool to eliminate the spurious correlations between geospatial context and trajectories. Extensive experiments on two real-world datasets verify that TrajCL markedly enhances performance in trajectory classification tasks while showcasing superior generalization and interpretability.