针对非线性时间序列，建立了小波神经网络预测模型，通过计算小波分解和小波级数，达到最优的逼近效果．经实例验证，该方法能有效地提高预测精度，避免了人工神经网络模型的固有缺陷.

In this paper, we build a wavelet neural network prediction model of time series. By computing the wavelet resolution and wavelet series, the best approximation effect is reached. It can be seen from the example that this method can effectively improve the prediction accuracy. The intrinsic defects of artificial neural network model are avoided.