A Generative Framework for the Creation of Multi-Attribute Geographically-Explicit Synthetic Population

A new generative framework is proposed to create multi-attribute, geographically-explicit synthetic populations. This framework addresses the challenge of reconstructing region-specific joint distributions from aggregated data, which is a foundational requirement for geo-simulation techniques like micro-simulation and agent-based modeling. The method utilizes a hierarchical diffusion-based approach to generate synthetic populations, including explicit home and work locations, trained on realistic region-specific joint distributions of multiple attributes.

Why it matters

This development significantly enhances the capability to create highly realistic synthetic populations, which are critical for robust geo-simulation and modeling across various domains. It allows for more accurate and granular analysis of complex systems by providing a foundational dataset that reflects both individual attributes and geographic context.

Key insights

• Existing methods struggle to reconstruct region-specific joint distributions from aggregated data for synthetic population generation.
• A hierarchical diffusion-based generative framework is proposed to overcome these limitations.
• The framework creates multi-attribute synthetic populations with realistic joint distributions and explicit geographic assignments (home and work locations).
• It is designed for geo-simulation techniques, including micro-simulation and agent-based modeling.
• The framework was applied across 50 U.S. states and Washington, D.C., generating a nationwide geographically-explicit synthetic population.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.12768