Research On Portfolio Optimization Based On Multi-Objective Evolutionary Algorithm

In recent years,thanks to the rapid development of economy and technology,the living standards of our people have continued to improve,and financial markets have gradually occupied an important position in economic activities.However,with strong blindness and herd mentality,ordinary investors have a lower ability to resist risks.It is of great significance to study how to effectively combine assets to find a more reasonable investment strategy.Multi-objective evolutionary algorithm(MOEA)is the main method to solve multiobjective optimization problem(MOP),It can obtain a set of equilibrium solutions that make each objective as optimal as possible.In the process of stock investment,investors hope to reduce risks while increasing returns,this is a typical multi-objective portfolio optimization problem.The content of this paper is mainly focused on the following two aspects: On the one hand,for the limitations of mean-variance model,the historical data is replaced by the return predicted by a heuristic functional link neural network.To reduce asset management difficulty and improve the computational efficiency,cardinality constraint is used to limit the size of portfolio,skewness is introduced to describe the asymmetry of the real return distribution,and then build a prediction based mean-variance-skewness cardinality constrained portfolio optimization model.On the other hand,for the practical problem,based on the NSGA-III algorithm,in the environment selection phase,a distance function that comprehensively considers convergence and diversity is introduced to replace the vertical distance,in the mating selection phase,tournament selection based on membership is introduced instead of random selection method,a coding scheme with asset information and two phase crossover and mutation operators are introduced to update asset type in a portfolio,feasible solution repairment operator is used to satisfy constraints and reduce the complexity of solving problem,and then propose a customized improved algorithm c-NSGA-III.Based on the practical stock data of the SSE A Share,a large number of comparative experiments and results analysis are performed,which fully verified the effectiveness of the improved model and improved algorithm.The experimental results show that the c-NSGA-III algorithm can optimize the problem solving process,which is better than NSGA-II and MOEA /D and has practical application value,it also provides an effective method for many objectives portfolio optimization problems.