Abstract:

Multi-objective particle swarm optimization based on a K-means guide selection strategy (KMOPSO) is proposed. A K-means algorithm based guide selection strategy is used to select K evenly located non-dominated particles from the archive in order to ensure the particles in the population move to the entire Pareto front and improve the diversity of solutions. A pruning method based on the nearest neighbour is adopted to control the size of the archive, while preserving the diversity of the archive. A mutation operator is presented to improve the exploration ability for preventing from premature. Simulation on five classical test functions indicates the feasibility of the proposed algorithm. KMOPSO can generate non-dominated solutions close to the true Pareto front and outperform non-dominated sorting genetic algorithm II and multi-objective particle swarm optimization with crowding distance in terms of the diversity of non-dominated solutions.

Key words: particle swarm optimization, multi-objective optimization, K-means algorithm