Constructing approximation models based on agent-based simulations by genetic algorithms

This paper presents an agent-based simulation approach to the construction of approximation models for macroscopic indices of economic and social systems(ESSs), in which a target ESS is modeled by a set of agents and the set is optimized by a genetic algorithm so as to behave similarly to the target ESS. Once an agent-based approximation model of an ESS has successfully been constructed, the ESS can be analyzed not only at macroscopic levels but also at microscopic levels through its agent-based simulations. To investigate the capability of the proposed approach, we make some experiments on the U-Mart system.