Predicting Pareto Dominance in Multi-objective Optimization Using Pattern Recognition

Abstract

A method of predicting Pareto dominance in multi-objective optimization using pattern recognition, and the framework of Pareto dominance classifier are proposed. A kind of Bayesian classifier is preliminarily implemented. It is used to predict Pareto dominance among the candidate solutions for various typical multi-objective optimization problems. The experimental data and analysis show that the predicted results can be used to recognize the non-dominated candidate solutions. Thus the prediction method of Pareto dominance can serve as an efficient approach for overcoming the curse of computation cost in solving complicated multi-objective optimization problems.