Towards Robust Trajectory Representations: Isolating Environmental Confounders with Causal Learning

arXiv:2404.14073v1 Announce Type: new
Abstract: 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 …

abstract acquisition arxiv behavior bridge capabilities causal confounding context correlations cs.ai cs.lg effects environmental gap geospatial human mobility modeling patterns pivotal robust studies trajectory type understanding