The impact of social network structure in particle swarm optimization for classification problems

Abstract

Particle Swarm Optimization (PSO) is a mechanism that involves several particles (solutions) interacting among each other to find the best solutions. It is a functional procedure by initializing a population of random solutions and searches its member by assigning random positions and velocities. The potential particle solutions are then flown through the hyperspace to get the optimum solutions. In this study, social network structure of PSO is incorporated into Artificial Neural Network (ANN) to investigate its learning efficiency. The results yield that the classification and convergence rates of ANN with ring structure (lbest) is better compared to the star social structure (gbest). These results are further validated by executing statistical significant test for better justification.