Prediction of high frequency data applied to the Bond Found Time Series: Neural approach

We develop forecasting models based on the neural approach for the forecasting of the bond price time series provided by the VUB bank and make their comparisons of the forecast accuracy with the class of the statistical ARCH-GARCH models. There is a limited statistical or computer science theory on how to design the architecture of the RBF networks for some specific nonlinear financial or economic time series, which allows the exhaustive study of underlying dynamics, and to determine their parameters. To illustrate the forecasting performance of these approaches the learning aspects of RBF networks are presented and the application is included. We show a new approach of function estimation for nonlinear time series forecasting model by means of granular neural network based on the Gaussian activation function and by incorporating a cloud concept in the RBF neural network. A comparative study shows that the presented approach is able to model and predict the high frequency data with reasonable accuracy and more efficiency than the statistical methods based on the ARCHGARCH methodology.