Genetic modelling of multivariate EGARCHX-processes: evidence on the international asset return signal response mechanism (Q1603562)

In this paper we compare a newly developed genetic algorithm (GA) to an extended MEGARCHX algorithm for modelling heteroscedastic vector processes. The competing algorithms are applied to studying the impact of the Japanese stock prices on the Finnish derivatives market. The algorithms have been initially tested on simulated data with known parameters. MEGARCHX-modelling of the Finnish stock returns and futures returns by traditional nonlinear (maximum likelihood) methods effectively captures the linear dependence and heteroscedasticity present in the series. However, the GA provides a powerful supplement to traditional econometric techniques.