Optimal Operators And Parameters Of The Genetic Algorithm In Solving The Portfolio Of Mean-VaR Kernel Estimation

VaR is a newly introduced method in the field of risk measuring. It is excellentbecause of its concise concept, suitability for supervising and comprehensivemeasurement. Historical simulation based on kernel estimation can be implementedon continuous returns. It has the advantage of not having to suppose the distribution,as well as more accurate in estimating and more estimation information.In this paper, VaR historical simulation based on kernel estimation is introducedinto Markowitz portfolio model, as a substitution to variance of the measurement ofrisk. An expected return and VaR historical simulation kernel estimation is built. Agenetic algorithm is to be used to solve the model considering the complexity of themodel.The genetic algorithm is an excellent globally and stochastically searchingmethod. It approaches the optimal solution through simulating the evolution ofcreatures in the nature. Sometimes it need special designing to get an efficient andaccurate solution to some special problems.During the period of solving the problem, operators and parameters must beselected optimally, because of the notable effects of different selection of operatorsand parameters. Thus a creative algorithm is given in the paper. It is an inside-outsidegenetic algorithm, with the operators and parameters of the inside being the codes ofthe outside and their efficiency of solving the inside being the fitness function of theoutside. The best operators and parameters of inside are selected while the solution ofthe outside is got.This method/algorithm can also be used in similar field which a geneticalgorithm is used to get solutions. After a pre-selection of operators and parameters, ahighly efficient and accurate solution can be reached.