Prediction of the smoothed monthly mean sunspot numbers by means of radial basis function neural networks

The Radial Basis Function (RBF) neural networks method is introduced and applied to the smoothed monthly mean sunspot number's (SMMSN) prediction for cycle 23 in this paper. Prediction methods are made respectively for predicting of SMMSNs for the next eight months by training the neural networks with different sets of data. A comparison of the SMMSN's predictions one to eight months in advance with the derived ones from the observational data for absolutely the most part of cycle 23 shows that this RBF neural networks method should be an applicable one for the mid-term solar activity forecast. A brief discussion give in the last section of this paper points out: (1) that the error of the prediction increases along with the time in advance, while for the prediction with an advanced time of <= 4 months the error can be controlled under 4. 8 and 38 %, and for 89 % of this kind of prediction the relative error is <= 15 %. (2) that size of the data set used for the training of the RBF neural networks would give an effect to the predicting ability of the prediction model.