Parameter calibration of a system dynamics model. A comparison of three evolutionary algorithms

Abstract

This research seeks to improve the parameter calibration process of a System Dynamics model. A movie release strategies" model has been developed in 2012 using a gradient-based optimization algorithm to estimate all the parameters. On this research, three modern optimization algorithms are initially compared using mathematical benchmark functions and then tested with the model to compare results. The tested algorithms are modifications of the Artificial Bee Colony algorithm, the Cuckoo Search and the Genetic Sampler. The results show that by using the Artificial Bee Colony algorithm, better performance is achieved in terms of speed and fitness. It is also shown how the optimization problem definition was improved resulting from a better optimization process.