Abstract: The prediction of colored noise is studied. A method of neural network prediction based on wavelet decomposition is proposed by combining the theory of wavelet analysis with the fundamental principle of neural network modeling and prediction. It has been verified that the prediction effect of the method is sound by predicting average sunspot numbers per year, which is a typical statistical model. Then the method is used in colored noise prediction. The possibility and effect of colored noise prediction is studied through changing the sampling rate for colored noise. The research results show that colored noise can be predicted, the prediction error decreases with the increase in the sampling rate and the precision of prediction of the method of neural network prediction based on wavelet decomposition proved to be better than the precision of prediction of neural network prediction.