A Modified Genetic Algorithm with Multiple Subpopulations and Dynamic Parameters Applied in CVaR Model

Simple genetic algorithm (GA) involves only one initial population with fixed genetic operational parameters selected in advance. This paper presents a modified genetic algorithm (MGA) with multiple subpopulations and dynamic parameters. In the new algorithm, genetic operations are carried out on each subpopulation separately and the operational parameters are changed dynamically. In the meanwhile, the genetic information of individuals is transferred among all the subpopulations by means of migration operator throughout the evolution processes. Furthermore, the modified genetic algorithm has been applied to conditional value-at-risk (CVaR) model to solve financial risk problems and the results indicate that the solution obtained from MGA is much better than the one obtained from a GA. To make this clearer, a numerical example is demonstrated for algorithm illustration.