The Atlantic ran a story based on an interview with William G. Kennedy, the principal investigator (PI) of the project that I have been lately working on. You can read more about the project following the links. In this post, I would like to share my experiences in generating a synthetic population using U.S. Census Bureau’s data, which is fed into our agent-based simulation.
I created a Jupyter notebook (Python 3) which shows how we generate our synthetic population on a sample of two counties in New York: Sullivan and Ulster. My goal was to create individuals (age and sex), living spaces (houses, workplaces, and schools), social networks (relationships between household members, school mates, and work colleagues), and also a giant connected road network for agents to travel.
We use three datasets from US Census to synthesize our population:
- Road Network
- Road data from Census TIGER shapefiles (primary, secondary, and residential roads)
- Commute flows
Finally here are some visuals of individuals moving on the road network, social networks in which they are embedded, and living spaces in a census tract: