The Performance of Artificial Bee Colony (ABC) in Structure Selection of Polynomial NARX Models

System Identification (SI) is a discipline of building a mathematical model of dynamic systems based on its input and output data. The process of SI is generally divided into structure selection, parameter estimation and model validation. This paper attempts to address the structure selection issue in SI, where the objective is to select the most representative set of regressors to represent the system. However, the selection process must obey the principle of parsimony, where the structure must be as small as possible, yet has the ability to represent the system well. We propose a binarized modification of the Artificial Bee Colony (ABC) algorithm to perform structure selection of a Nonlinear Auto-Regressive with eXogenous (NARX) model on a Direct Current (DC) motor. We compare this implementation with the Binary Particle Swarm Optimization (BPSO) algorithm in terms of solution quality and convergence consistency. The results indicate that the ABC algorithm excelled in terms of convergence consistency with similar solution quality to BPSO algorithm