Neural network prediction of solar activity

Abstract

The neural network technique is used to analyze the time series of solar activity, as measured through the relative Wolf number. First, the embedding dimension of the time series characteristic attractor is obtained. Second, after describing the design and training of the net, the performance of the present approach in forecasting yearly mean sunspot numbers is favorably compared to that of conventional statistical methods. Finally, predictions for the remaining part of the 22d and the whole 23d cycle are presented.