Parameter tuning in agent based simulation

This study was carried out to solve the parameter tuning problem of complex systems that are modelled in an agent-based modelling and simulation environment. Complex systems that are modelled need a good set of parameters to be able to demonstrate the realistic and desired behaviour. This process is an optimization problem that must be solved for complex systems with large parameter spaces. This study presents a tree step hybrid parameter tuning approach for agent-based models and simulations. In the first step, the problem is defined; in the second step, parameter tuning process is carried out using Genetic Algorithms, Particle Swarm Optimization, Artificial Bee Colony and Firefly Algorithms, which are meta-heuristic algorithms. Again, meta-heuristic algorithms used in the second stage have been observed in this thesis that critical parameter values directly affect the performance of the algorithms. Therefore, adaptive meta-heuristic optimization algorithms which are adapted to the problem are developed using the online adaptive parameter control technique. These algorithms are also used to adjust the model parameters. In the third step, which is the control phase, the algorithm parameters obtained by the online adaptive parameter control method and the parameter values of the model from the used meta-heuristic algorithms are manually provided and tested. This approach is compared with the algorithm performance of the problems tested in 3 different models and 4 different numerical test functions.