非线性时间序列的小波网逼近及其预报

The approximation and predication of non-linear time series by using wavelet neural networks

结合神经网络和小波的优点，建立了一种新型的多维小波网的网络模型，研究了多维小波网逼近非线性时间序列的收敛性质及收敛精度，并给出了相应的数学证明，同时将之应用于山猫数据的拟合及预报，仿真结果说明该方法的可行性。

The authors establish a new multi-dimensional wavelet network model by combining neural network and wavelet characteristics. They also study the approximation properties and accuracy of wavelet network to approximate non -linear time series of dimensional wavelet network and give the mathematical proof, at the same time, apply the net to approximate and predicate Shanmao datas, the result reveals that the algorithm is feasible.