A Clustering-Based Hybrid Particle Swarm Optimization Algorithm for Solving a Multisectoral Agent-Based Model

This paper presents the new Clustering-Based Hybrid Particle Swarm Optimization (CBHPSO) algorithm. This algorithm was designed to solve biobjective optimization problems and it was used for finding trade-offs in a multisectoral agent-based model of trade interactions. This model includes multiple interacting agent enterprises belonging to different economic sectors. At the same time, the values of the control parameters of this stochastic multiagent system (MAS) need to be optimized. Therefore, CBHPSO has been developed and aggregated with this MAS by means of the objective functions. The main feature of the CBHPSO algorithm consists in the use of clustering techniques, such as the k-means algorithm, to form subsets of non-dominated solutions shared among the swarm particles in each cluster. The values of the performance metrics employed for CBHPSO and other well-known multi-objective evolutionary algorithms (SPEA2, NSGA-II, FCGA and BORCGA-BOPSO and MOPSO) were compared. As a conclusion, it was found that the velocities of decision variables in the particle swarm depend on specific non-dominant solutions of the clusters involved. It can be said that the main advantage of CBHPSO lies in the quality of the Pareto front approximation. Thus, it was demonstrated that the CBHPSO algorithm can be applied to search for improved characteristics of the employed MAS.