Several recent studies have applied neural network models to the analy
sis and prediction of time series behaviour of solar activity. These s
tudies have in general used overall measures of error magnitude to rat
e prediction accuracy. In this study, we concentrate on prediction of
solar activity and demonstrate the tendency of neural networks to gene
rate delayed predictions of specific features in the data. The necessi
ty of recognising delayed predictions is discussed and a new training
algorithm, based on a modified approach combining a genetic algorithm
with back-propagation of errors, is proposed to counteract the problem
. (C) 1998 Elsevier Science B.V. All rights reserved.