PREDICTION OF SOLAR AND GEOMAGNETIC-ACTIVITY DATA USING NEURAL NETWORKS

Accurate predictions of the future behavior of the solar activity cycl e have been sought for many years, Several classes of prediction appro ach have been proposed, with many variations in each class, and have a chieved varying degrees of success, However, considerable room for imp rovement still remains, Artificial neural network models enjoyed a res urgence in popularity as prediction tools during the late 1980s, as a consequence of the discovery of the back propagation of errors learnin g algorithm, Initial investigations have been carried out into their p otential for predicting solar activity (e.g,, Koons and Gorney, 1990; Williams, 1991; Macpherson, 1993a, b). In this paper, we investigate i n detail the effect different neural network architectures and learnin g parameters have on the prediction accuracy of various networks train ed on smoothed monthly sunspot and solar 10.7-cm flux data, The import ance of obtaining the best generalization capability of a neural netwo rk is stressed, Prediction of the geomagnetic aa index is also conside red, Finally, in order to validate the usefulness of this technique, t he results are compared with a variant of the well-established McNish and Lincoln method (McNish and Lincoln, 1949) and are found to be supe rior in terms of prediction accuracy.