MODELOVANIE VOLATILITY A PREDIKČNÉ MODELY VYSOKOFREKVENČNÝCH FINANČNÝCH DÁT: ŠTATISTICKÝ A NEURONOVÝ PRÍSTUP
Volatility modelling and the forecasting models of high frequency financial data: Statistical and neural approach
Languages of publication
In the article we first introduce asymmetric response of equity volatility to return shock and then the effect of good and bad news to volatility for empirical time series of EUR/USD (EUR currency against US dollar) exchange rates in the pre-crisis period, during the crisis and the post-crisis period. We found that GARCH-class models with normal errors are not capable to capture fully the leptokurtosis in empirical time series, while Student´s t and GED errors provide better description for the conditional volatility. Then, we alternatively develop forecasting models based on the ARIMA/GARCH methodology and on the neural approach. In the direct comparison between statistical and neural models, the experiment shows that the neural approach clearly improve the forecast accuracy.
133 – 149
Publication order reference