A dual-operator strategy for a multiobjective evolutionary algorithm based on decomposition

Abstract

Evolutionary Algorithms (EAs) are a kind of population based on optimization method by adopting survival of the fittest rules. The performance of EAs can be greatly improved by appropriate genetic operators, so how to select an appropriate genetic operator is a key issue. In order to solve this problem, some genetic operators are mixed to use with a certain probability to improve their spatial search capabilities. However, it is difficult to solve most complex multi-objective problems (MOPs) based on a certain probability value. In this paper, under the concept of co-evolution and Duty Ratio, we built a genetic operator based on Differential Evolution (DE) and Simulated Binary Crossover (SBX), and the adjustment of Duty Ratio parameter is learned based on the historical used times of DE and SBX. Under the framework of Multi-objective Evolutionary Algorithm Based on Decomposition (MOEA/D), we propose our dual-operator strategy (DOS) based on learning strategy, namely MOEA/D-DOS. We compared MOEA/D-DOS with other six versions of multi-objective EAs, and the final result showed that MOEA/D-DOS has achieved the better results.