How to derive spatial agents: A mixed-method approach to model an elderly population with scarce data

Information about the spatial patterns of residents is essential, especially when elderly people are involved, as their action range is confined to their residential location. Since knowledge about patterns of elderly people in cities is limited, this paper formulates steps for the initialisation of an agent-based model, combined with different data sources. The first step is to identify different types of elderly people using cluster analysis, and then the clusters are expanded into agent typologies with behaviour rules, which form the basis for an artificial population. The clusters are derived based on survey data and then analysed and modified using insights from census data and expert interviews. The agents' relocation behaviour is estimated based on literature research, expert interviews and a survey. The spatial information of the agents is added with a spatial microsimulation. The resulting artificial population presents the real population well and can be used in an empirically based data-driven agent-based model.