Abstract:A performance evaluation study is implemented between the methods of Genetic Algorithms with Floating Point representation and some traditional optimization methods, in the task of estimating the parameters of a GARCH (1,1) Normal process, using artificial data obtained by simulation. The results show that the approximate solutions obtained by means of Genetic Algorithms present a better stability and precision with respect to the traditional optimization methods. The choice of the initial point in numerical optimization methods is not a critical condition in the use of Genetic Algorithms as a method to find the solution. Finally, Genetic Algorithm method is illustrated in the finding of the solution of the vector of parameters of the likelihood function of a GARCH (1,1) t-Student model, using data of rates of exchange returns of the Sol against to the Dollar.
|
