摘要：Multi-Step ahead prediction of a chaotic time series is a difficult task that has attracted increasing interest in recent years. The interest in this study is the development of nonlinear neural network models for the purpose of building short term and long term ahead prediction of monthly sunspots chaotic time series. The solar activity has a measure effect on earth, climate, space weather, satellites and space missions and is a highly nonlinear time series. Such problems exhibit a rich chaotic behavior. In this study the authors compared the performance of three neural network configurations namely a Multilayer Perceptron (MLP), Self Organized Feature Map (SOFM) and Focused Time Lagged Recurrent Neural Network (FTLRNN) for 1, 6, 12,1 8 and 24 months ahead prediction with regards to various performance measures Mean Square Error (MSE), Normalized Mean Square Error (NMSE) and regression (r). The standard back propagation algorithm with momentum term has been used for all the models. The various parameters like number of processing elements, step size, momentum value in hidden layer, in output layer the various transfer functions like tanh, sigmoid, linear-tan-h and linear sigmoid, different error norms L1, L2, L3, L4, L5 to L8, different combination of training and testing samples and epochs variation are exhaustively experimented for obtaining the proposed robust model for the short term and long term ahead prediction. The results suggests that the FTLRNN Model can help in improving the prediction accuracy.
