Learning Demographic-Conditioned Mobility Trajectories with Aggregate Supervision

2026-03-03 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Social Sciences & Behavioral Studies · Depth: Expert, quick

Summary

ATLAS is a weakly supervised approach designed to generate demographic-conditioned human mobility trajectories, addressing the common lack of demographic labels in existing trajectory datasets. It utilizes individual trajectories without labels, region-level aggregated mobility features, and region-level demographic compositions from census data. The model trains a trajectory generator and fine-tunes it to ensure simulated mobility patterns align with observed regional aggregates, while explicitly conditioning on demographic information. Experiments using real trajectory data with demographic labels demonstrate that ATLAS significantly enhances demographic realism compared to baseline models, achieving a JSD reduction of 12% to 69%, and substantially narrowing the performance gap to strongly supervised training methods. The research also includes theoretical analyses explaining the efficacy of ATLAS, highlighting factors like demographic diversity across regions and the informativeness of aggregate features.

Key takeaway

For research scientists developing mobility models for public health or social science, ATLAS offers a robust method to incorporate demographic heterogeneity even when individual trajectory data lacks explicit labels. You should consider integrating region-level aggregate mobility features and census data to significantly improve the demographic realism of your simulations, especially when strong supervision is unavailable. This approach can enhance the accuracy of predictions and policy recommendations.

Key insights

ATLAS generates demographically realistic mobility trajectories using weak supervision from aggregate regional data.

Principles

• Aggregate data can infer individual attributes.
• Demographic diversity improves model performance.

Method

ATLAS trains a trajectory generator, then fine-tunes it to match simulated mobility with observed regional aggregates, conditioned on census-derived demographics.

In practice

• Use census data for demographic conditioning.
• Leverage regional aggregates for weak supervision.

Topics

• Mobility Trajectories
• Demographic Modeling
• Weakly Supervised Learning
• Aggregate Supervision

Code references

• schang-lab/ATLAS

Best for: Research Scientist, AI Researcher, AI Scientist, Machine Learning Engineer

Related on AIssential

• GPS-Enhanced Tourist Mobility Modeling with Seasonal Spatial Priors and LLM-Based Activity Chain Generation — The framework combines GPS data, demographics, and LLMs to simulate realistic, demographically-aligned tourist mobility for urban planning.
• The Interplay Between Interpolation and Aggregation in Regression: Optimal Sample Complexity — The γ-graph dimension characterizes learnability, and a median-based aggregation of three interpolating hypotheses is optimal.
• C2FL: Clustered Continual Federated Learning under Spatial and Temporal Drift — C2FL enables robust federated learning in mobile, privacy-sensitive systems by combining spatial clustering with temporal drift adaptation.
• Causal Decomposition Analysis with Synergistic Interventions: A Triply-Robust Machine Learning Approach to Addressing Multiple Dimensions of Social Disparities — An extended causal decomposition analysis evaluates synergistic effects of multi-domain interventions using a triply robust machine learning estimator.
• S2MAM: Semi-supervised Meta Additive Model for Robust Estimation and Variable Selection — S²MAM uses bilevel optimization for robust semi-supervised learning, identifying variables and updating similarity for interpretable predictions.
• TrajGenAgent: A Hierarchical LLM Agent for Human Mobility Trajectory Generation — TrajGenAgent's hierarchical LLM-agent framework generates realistic human mobility trajectories by decoupling semantic planning from spatiotemporal grounding without fine-tuning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.