Heuristic Data Merging for Constructing Initial Agent Populations

In this paper, we explore an approach for developing an initial agent population that is suitable for integrating two component agent based models, representing conceptually the same agents. For some models the structure of the initial population is an important aspect of the model. When integrating two (or more) models that represent the same agents, we require a single integrated agent population (or unique mappings between the two populations). Obtaining such is not straightforward if we wish to preserve important structural characteristics of the component populations. We describe here a methodology inspired by work in constructing synthetic populations which are structurally similar to a real population. The approach uses the Iterative Proportional Fitting Procedure (IPFP) to combine two different data sets in a way that preserves the structure of each. We apply our approach to a specific case study and evaluate the quality of the resulting integrated population.