Sensitivity Analysis of a Parallel Particle Swarm Optimization Clustering Algorithm for Multi-objective Optimization

Parallel bio-inspired algorithms have been successful in solving multi-objective optimisation problems. In this work, we discuss a parallel particle swarm algorithm with added clustering for solving multi-objective optimisation problems. The aim of this work is to perform sensitivity analysis of the parallel particle swarm algorithm. We need to see how the added parallelism improves the overall execution time. Also, looked at the effect of different strategies for population initialisation (such as mutating current set of leaders, random population and lookup in archive for nearest points using geometric calculation). The results show that using different migration frequencies for scattering reduced the overall overlap between processors. Results regarding how clustering and gathering affect performance metrics are also reported.