Self-Adaptive Evolutionary Computation Methods for GARCH Model Optimization

Abstract

This paper deals with the parameter estimation of popular GARCH(1,1) model using an effective hybrid evolutionary computation (EC) based method. These parameters are estimated by maximizing the nonlinear log-likelihood function. Further, in this study, the effectiveness of the EC based methods is verified through a comparative study with that of popular optimization methods such as Marquardt, BHHH etc. incorporated in the ready made software packages like MATLAB, EViews and Excel. Here, two EC-based methods, one a fast evolutionary programming method and the other one, a hybrid evolutionary computation method are considered for exploration and subsequent discussion.