A hybrid intelligent system for financial time-series forecasting

We present a hybrid intelligent system for financial forecasting that combines neural networks with econometric GARCH models for volatility. We show how this flexible modelling framework can accommodate most of the statistical features observed in financial time-series (nonlinearities in mean, asymmetric GARCH effects and non-gaussian errors). We analytically discuss several strategies for the specification of the mean and variance components of the model by means of sequential statistical tests and examine variations of the standard testing framework that are robust to misspecification of the underlying distribution. We demonstrate various aspects of the model-building strategy by employing NN-GARCH to predict the conditional distribution of daily returns on major international stock indexes.