摘要
提出一种用正交尺度函数代替RBF网络中的激活函数的小波网络，给出相应小波网络学习算法；并以我国1978～2001年的税收数据为样本进行税收模拟预测，预测结果表明，该模型预测误差低于普通BP网络。
In this paper, a wavelet-based neural network is given. The structure of this network is similar to that of the radial basis function （RBF） network, except that here the radial basis functions are replaced by orthogonal scaling functions. Furthermore, the paper proposes the algorithms for the wavelet network. Finally, a tax forecasting based on the sample of tax data （1978-2001） in China is presented, which shows that the forecasting error is less than that of BP network.
出处
《系统工程理论方法应用》
北大核心
2006年第1期54-56,70,共4页
Systems Engineering Theory·Methodology·Applications
关键词
小波神经网络
正交小波
税收预测
wavelet neural network
orthogonal wavelet
tax forecasting
