Agent-based models are a powerful tool for understanding the ways in which human populations may react to changes in their environment, social structure, or ruling body policies. Agent-based models allow researchers to measure the impacts of policies such as (1) closing schools, workplaces, and churches, (2) travel restrictions (e.g., roadblocks, border closings), and (3) distribution of relief supplies.

Agent-based models assign behaviors and activities to agents (i.e., individuals) within the population being modeled and then allow those agents to interact with the environment and with each other in complex simulations. To accurately predict population responses, agent-based models depend upon geospatially and demographically accurate population databases.

This paper describes lessons learned in generating a synthesized population database that represents the entire U.S. household population. An iterative proportional-fitting method was used to generate a synthesized, geospatially explicit human-agent database. The synthesized agent database locates individuals within households, places the households relative to census geography, and provides demographic attributes consistent with the census data. The demographic data supports modeling of interactions through social networks. This database was used for "what-if" studies of government responses to influenza epidemics.

This paper describes extending this methodology and these processing steps to generate synthesized agent databases in other parts of the world. We present the results of an effort to create a realistic synthesized agent database for Mexico, which could then be used in agent-based models to study the effects of potential economic and health policy on cross-border migration. The resulting synthesized population database is geospatially explicit and accurately represents the actual household population of Mexico. As more census microdata become available, synthesized population databases such as the one described in this paper become possible for more countries.

Brandon Bergenroth

Affiliation

Refine Results

Generating Human Terrain for Agent-Based Models