Building Threshold GARCH Models Of Time Series Based On Genetic Algorithms And Its Study And Application

In the analysis of financial time series, the GARCH Model is suitable for depicting the conditional heteroscedasticity of financial time series and the threshold model (threshold autoregressive or threshold ARMA model) is able to quite accurately describe the non-linear rules of series. The GATCH model in this paper is built by the combination of the advantages of these two. Since a series of coefficients are needed by the structural parameter of threshold GARCH model, the effective way is to use the genetic algorithms to search the structural parameter space. The choice of genetic algorithms is a simulation of the law of survival of the fittest in the evolution of species. Because it's bringing about quite ideal effects in processing large number of discrete and effective avoiding falling into local optimal solution, it has a much boarder searching scope, stronger adaptability, higher efficiency and better results and can overcome some shortcomings of the traditional H. Tong method, D. D. C. method and Local searching methods.By the comparison of the results of empirical research of the Shanghai Composite Index through a selected total of 1945 launch dates from January 2nd, 2003 to January 10th, 2011, this paper has discovered that the threshold GARCH model fitting and prediction accuracy is of a slight advantage in the processing of data and therefore is quite suitable for depicting non-linear laws. In addition, as the existence of random makes the constructed model different, rich and varied, it is easy for the decision-maker to select the appropriate model to perform time series analysis and financial explanation.