It's a focus for researchers to forecast the nonlinear data got from the reality world system. Many methods based on nonlinear models are proposed. In this paper, a forecast model based on the wavelet analysis and the neural network is proposed. Firstly, the wavelet decomposing and reconstruction should be made on the time series studied, and then neural network forecast method is applied on all the time series got from the analysis, the final forecast result is the sum. The daily data of a river is nonlinear data, and the effectiveness of the model is verified by applying the method to the daily data of a river.