HYBRID NEURAL NETWORKS AS A NEW APPROACH IN TIME SERIES FORECASTING

Nowadays there exists a various number of ways for time series prediction. Econometric analysis includes mainly ARIMA and GARCH models. However, in recent years, as computers become more and more the part of our lives, methods of machine learning have been used; among them artificial neural networks. This work combines these two independent and different prediction models into one combined model for time series prediction as it is an assumption that the combination could achieve better prediction results than the individual models. This new approach is then applied to exchange rates data to illustrate the sense of using this hybrid modelling technique.