A Study of Six Neighborhood Selection Methods for Multiobjective Neighborhood Field Optimization Algorithm

摘要

Neighborhood field optimization (NFO) algorithm is an evolutionary computation method which focuses on the power of neighborhood field to solve global optimization problems. NFO can also be able to solve multiobjective optimization problems (MOPs). This paper attempts to study the effect of neighborhood for multiobjective neighborhood field optimization (MNFO) algorithm. The study collects six neighborhood selection methods. The baseline of the six methods is random selection method. Two of them are based on solution dominance relation of the non-dominated sorting genetic algorithm II (NSGA-II). Two of them are based on fitness computation of the improving strength Pareto evolutionary algorithm (SPEA2). The remaining one is based on region-based selection method of the Pareto envelope based selection algorithm (PESA-II). Simulation is performed on twelve unconstrained test functions. The results show that random selection method is worse than other methods. Neighborhood selection based on SPEA2 shows better results than other methods.