Generating a Synthetic Population of Agents Through Decision Trees and Socio Demographic Data

Abstract

[Abstract]: Agent based models (ABM) are computational models employed for simulating the actions and interactions of autonomous agents with the objective of assessing their effects on the system as a whole. They have been extensively applied in social sciences because ABM simulations, under different running conditions, can help to test the implications of a policy intervention or to observe the population dynamics in different scenarios. We have developed an ABM to model how citizens behave with respect to superblocks, i.e., a type of social innovation where the urban space is reorganized to maximize public space and foster social and economic interactions while minimizing private motorized transports. In this model, the main entity is the citizen agent, so we must acquire personal attribute information to calibrate, validate, and apply the model to test different policy scenarios. Two main data sources were used to derive this information: census data and a survey. However, both were insufficient to generate a realistic population for the model. In this work we present how decision trees were used to generate a synthetic population using both types of data sources.