Research On Differential Evolutionary Multi-objective Optimization Algorithms Based On MOEA/D

An Evolutionary multi-objective optimization algorith,is that whose main task is to deal with multi-objective optimization problems by evolutionary computation,has become the focus of research in the intelligent optimization field.There are many multi-objective optimization problems in real life.Sometimes the different goals of these problems are mutually reinforcing,sometimes they are conflicting.The purpose of solving these problems is to find a compromise solution set.Currently,evolutionary multi-objective optimization algorithm has been widely applied to portfolio investments,route planning,network routing,data mining and so on.In theoretical research,the important research targets of evolutionary multi-objective optimization algorithm is to improve the convergence and distribution of the algorithm.This paper is based on MOEA/D.It proposes IMOEA/D and IMOEA/D-GSTS by improving the convergence and distribution of MOEA/D.The beginning of the paper,introduces the background and significance of evolutionary multi-objective optimization algorithm.Then,it illustrates some important concepts of evolutionary multi-obj ective optimization algorithm,indicators of quality of metrics algorithm,the difficulties of algorithmic research and the main design goal of algorithmic.Finally,this article proposes IMOEA/D,which is based on MOEA.IMOEA/D improves the mutation of MOEA/D by using bidirectional research strategy,so that mutation operation produces two individuals can be produced during each mutation operation.Therefore,IMOEA/D broadens the searched space of algorithms,and it also accelerates the convergence rate of algorithms.At the same time,instead of the selection strategy of MOEA/D,IMOEA/D use the selection strategy that is based on governable amount,so that more outstanding individuals takes part in the evolution of next generation.The experiment shows that bidirectional research strategy and selection strategy that is based on governable amount improve the convergence and distribution of the algorithm.In order to further improve the convergence and distribution of the algorithm.This paper also proposes to use each individual acceleration as mutation operators instead of mutation operators that is based on gaussian distribution.This strategy is based on the gravitational search algorithm.In addition,this strategy can take full advantage of the gravitational search algorithm,which are exploitation of early evolution and exploration of late evolution.In order to avoid trapping in local optimum,this paper regards the output of genetic algorithm as the input of tabu search and effectively uses local search capabilities through the tabu search algorithm.The two optimization strategies of IMOEA/D effectively improve the performance of the algorithm.Therefore,the paper proposes IMOSEA/D,which is based on the gravitational search algorithm and the tabu search algorithm.At the end of the paper,it summaries some results and shortcomings of the paper.