A 2-stage Artificial Neural Network predictor with application to financial time series

Il volume appartiene alla collana "Studies in Fuzziness and Soft Cmputing" edito da Kacprzyk J..

The purpose of this work is to develop a neural network forecasting system based on a 2-stage approach. In the first stage, some networks are realized in order to consider different forecasting model specifications, each being referred to a particular forecasting scenario. In the second stage, the different outputs (predictions) coming from each first stage forecasting model are suitably combined to generate a final output, whose quality level is expected to be higher. Different time series are used and the corresponding results critically analyzed and compared.