Fitting stochastic volatility model through genetic algorithm

Abstract

Abstract : The financial time-series data of many agricultural commodities show heteroscedasticity. So, the behaviour of prices of such commodities is fundamental to policy makers. One novel approach for modelling the volatile data sets is the promising methodology of Stochastic volatility (SV) model. SV model assumes the volatility to be an unobservable state variable following some latent stochastic process. In the present study, we aim to devise a procedure for estimation of parameters of SV using Genetic algorithm. Subsequently, the unobservable volatility is estimated using Kalman filter. For illustration purpose, the All-India data of month-wise total export of Basmati rice is considered. Comparative study to infer about the utility of SV model is also carried out by calculating various measures of goodness of fit and forecasting performance of the fitted SV model and GARCH model. Finally, it is concluded that SV model has performed better than GARCH for the data under consideration.