多步预测的小波神经网络预报模型

采用小波分析与人工神经网络相结合的方法，进行了一种新的非线性长期预报模 型研究。结果表明，所构造的小波神经网络多步预测模型对历史样本具有很好的拟合效果， 独立样本的预报效果也令人满意。并且，该预报模型具有误差收敛快，预报时效长以及建 模所需阶数低等显著优点，具有很好的研究前景。

A Multi-Step Prediction Model of Wavelet Neural Network

A new nonliner long-range prediction model is proposed based on the wavelet transform and wavelet neural network in the paper. Results indicate that multi-step prediction model of wavelet neural network is well available for both the fitting and the forecast. At the same time, the new prediction model shows great promise in the research of long-range prediction due to its fast convergence of the errors, long period of validity and low order to build prediction model.