EFFECTS OF DIFFERENT POPULATION SYNTHESIS MODELS IN THE EPIDEMIOLOGICAL SEIR (SUSCEPTIBLE-EXPOSED-INFECTED- AGENT-BASED MODEL

The advent of population synthesis has paved the way for the development of urban microsimulation studies, particularly in epidemiology. Thus, models must utilize synthetic populations that accurately match the actual population of a region, which is essential for large-scale agent-based modelling (ABM) efficiency and validity of large-scale ABMs. Although multiple population synthesis models exist, little is known about whether the errors generated during synthetic population generation can propagate through the simulation model and affect outcomes; this is the aim of this study. The SEIR(SusceptibleExposed-Infected-Released)-ABM is required to simulate the dynamics of disease spread independently using the resulting synthetic population generated through Iterative Proportional Fitting (IPF) and Markov Chain Monte Carlo (MCMC). The performance metrics used to assess the goodness of fit of IPF showed a high percentage of error compared with MCMC, specifically for the type of commuting and spatial locations of synthetic agents. In addition, a slight variation in the workingclass group between the population models resulted in different inferences in the disease spread prediction. For instance, the average peaks of new infections in the IPF and MCMC models were 42.2 and 34.7, respectively. The simulation results indicated that inherent errors in social network attributes concede to the mispredictions of epidemiological simulation outcomes.