Abstract
Adaptive stochastic search methods are expected to lead to "optimal" or "near-optimal" configurations of a simulation model as they manage to escape from sub-optimal (local) solutions. In that sense, they provide an automated "optimization" approach that adapts the parameters of a model in order to handle uncertainty that arises from stochastic elements in either the environment (process noise/concept drift) or the objective function evaluation process (observation noise) and improves the performance of the model. The paper reviews the fundamentals of adaptive stochastic search methods and explores their behavior for the adaptation of the parameters of a steelworks model. Experimental results illustrate the effectiveness of the methods, and particularly of swarm intelligence in this task.