Improved Evolutionary Multi-objective Optimization Algorithm Based On Decomposition And Its Application

Multi-objective evolutionary algorithm through decomposition(MOEA/D),as a typical algorithm in EA based multi-objective optimization algorithms,has the characteristics of strong search ability and independent of specific problems and,nowadays,it is becoming a fast developing and better optimization method in solving multi-objective optimization problems.However,MOEA/D and its variants generally have the following problems:(1)the global and local search ability of the algorithm is low;(2)the distribution of solutions is uneven;(3)the performance of dealing with discrete problems is poor.In view of these problems,the paper proposes a Multi-objective Evolutionary Algorithm through Decomposition based on Monte Carlo and Self-Adaption difference operator(MOEA/D-MC-SA),which is further improved and used to deal with knapsack problems.Aiming at shortages such as the low global and local search ability of algorithm,the uneven distribution of solution and the poor performance of dealing with discrete problem in multi-objective evolutionary algorithm through decomposition,this paper proposes the corresponding improvement methods and the main works includes:1.Multi-objective Evolutionary Algorithm through Decomposition based on Monte Carlo and Adaption difference operator(MOEA/D-MC-SA)is proposed.The main contribuation of the proposed algorithm:(1)Monte Carlo method is used to generate initial population in MOEA/D-MC-SA rather than random initialization used in most of EAs,which can avoid low search ability caused by the uneven distribution of solutions.(2)an adaptive difference volutionary operator is designed,which can dynamically adjust the scaling factor F and the crossover probability CR so as to improve the local search ability.The experimental results show that MOEA/D-MC-SA has good performance compared with other four Multi-Objective Evolutionary Algorithms on several test problems.2.A MOEA/D-MC-SA-Discrete algorithm for solving multi-objective knapsack problem is proposed,which is an improved version of MOEA/D-MC-SA.The main contribuation of the proposed algorithm: It can solve discrete multi-objective optimization problems,and it mainly solves the poor performance of the MOEA/D-MC-SA algorithm in dealing with discrete problems.Since discrete multi-objective optimization problems are verycommon in life,the multi-objective knapsack problem is a typical problem in discrete multi-objective optimization problems.Therefore,applying the improved MOEA/D-MC-SA-Discrete algorithm to the multi-objective knapsack problem has practical significance.