Regresar

Multi-objective genetic algorithms: Are they useful for tuning parameters in agent-based simulation?

Abstract:

One of the main issues, when an agent-based simulation (ABS) is performed, is the existence of a lot of parameters that should be tuned to adequately represent the dynamics of the system. Recently, Multi-Objective Genetic Algorithms (MOGAs) have been used to parameter tuning in some areas obtaining interesting results. In this paper, we studied the usefulness of a MOGA for parameter tuning in a well-known model presents in the NetLogo library about a colony of ants forages for food. We considered different scenarios (numbers of ants) while we used a MOGA to obtain solutions that represent the best combinations of the two parameters which minimize the emptying time of each food source. Simulations were carried out by using both optimized and random parameter values. The computational results indicate that in five out seven scenarios, the difference in average times between both types of parameters values used in simulation is statistically significant.