Generating a Located Synthetic Population of Individuals, Households, and Dwellings

Some of the new approaches in urban modeling, such as multi-agent systems (MAS) or activity-based models (ABM), require inputs in the form of disaggregated individual data. But for privacy protection reasons, such data is seldom available at this level. One way to get around this obstacle is to generate a synthetic population. This paper presents a method for generating a population from fully aggregated socio-demographic and geographic data. Based on French examples, this method can be reproduced anywhere in the country providing a relevant linkage between the characteristics of agents and those of urban spaces. The proposed method is subdivided into two steps. First, a population of agents belonging to households, as well as of households ascribed to housing units, is generated from the socio-demographic data. Second, this population is located by assignment to the buildings generated from the geographic data. Testing and validating the method on three French cities (Besancon, Strasbourg and Lille) generates useful results but also some difficulties, particularly for certain population categories. Ultimately, we obtain a realistic three-dimensional database of the study area where agents and spaces are represented and realistic individual information can be mapped and used to model the behavior of agents through MASs or ABMs.