Generation of Pareto Optimal Solutions Using Expected Improvement and Generalized Data Envelopment Analysis
Yun Yeboon, Nakayama Hirotaka
Evolutionary optimization methods, for example, genetic algorithms and particle swarm optimization have been applied for solving multi-objective optimization problems, and have been observed to be useful for generating Pareto optimal solutions. In order to generate good approximate and well-distributed Pareto optimal solutions with a small number of function evaluations, this paper suggests a new recombination method by utilizing expected improvement and generalized data envelopment analysis in a real-coded genetic algorithm. In addition, the effectiveness of the proposed method will be investigated through several numerical examples, by comparison with the conventional methods.