Comparison of neural network and McNish and Lincoln methods for the prediction of the smoothed sunspot index

In this paper we propose a comparison between two methods for the problem of long-term prediction of the smoothed sunspot index. These two methods are first the classical method of McNish and Lincoln (as improved by Stewart and Ostrow), and second a neural network method. The results of these two methods are compared in two periods, during the ascending and the declining phases of the current cycle 22 (1986-1996). The predictions with neural networks are much better than with the McNish and Lincoln method for the atypical ascending phase of cycle 22. During the second period the predictions are very similar, and in agreement with observations, when the McNish and Lincoln method is based on the data of declining phases of the cycles.