Estimation of macroscopic parameter in agent-based pandemic simulation

Simulations of epidemic spread is an appealing approach to design intervention programs against influenza epidemic. Agent-based approach is particularly useful, since it enable us to administrate and evaluate effectiveness of various intervention measures in the simulated city. On the other hand, observation date are obtained as macroscopic information. Hence, we need to extract some macroscopic information, to examine the validity of the simulation result. The reproduction number, which indicates how many new infected persons are produced due to one infected person, is often used in epidemiology. This quantity is directly related to coefficients in differential equations in macro simulations, whereas it is non-trivial in agent-based simulations. In this paper, we demonstrate the estimation of the reproduction number from an agent-based simulation result, by assimilating the number of infected persons yielded by an agent-based simulator to the SEIR model, which is a representative macro simulation model.