Data Driven Simulation Modeling for Mobile Agent-based Systems

Simulation modeling is generally knowledge-driven, where the models are developed based on knowledge about how the systems under study work. In this paper, we present a data-driven simulation modeling approach where the simulation model is learned from behavior data observed from the systems under study. As a first step of this work, we focus on mobile agent-based systems and develop a data-driven simulation modeling framework for mobile agent-based systems. The developed framework includes a specification of a readable solution space for mobile agent-based models, and a genetic algorithm-based solution space search strategy. The results show that not only are the discovered models easy to interpret, but they can be used to infer the existence of models that exhibit certain behavior.