Abstract: Nonlinear prediction and RBF (Radial Basis Functions) neural network prediction of EEG(Electroencephalogram) signal in a Liley model are studied by phase graph and NLCP (Nonlinear Cross-Prediction).It concluded that:1) RBF neural network prediction is better than nonlinear prediction;2) NLCP method is adaptive to time series with strong periodic components;3) support the exist of chaos in EEG signals.

Key words: Liley model, Electroencephalogram, Nonlinear Cross-Predication, Radial Basis Functions neural network prediction, chaos