基于小波神经网络的时间序列流数据的研究

Research for Model of Time Series Streaming Based on Wavelet Neural Network

利用多分辨分析方法,结合小波分析和神经网络思想构建一种新型的神经网络模型———小波神经网络,解决了传统神经网络中隐层节点数难以确定的问题。通过对股票的预测,说明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,by utilizing the method of multi - resolution analysis and combining the theory of wavelet and neural network, build a new model of neural network-wavelet neural network prediction model of time series, and solve the difficult problem of the hidden node decision in traditional neural network. The example of stock prediction testifies that this method can effectively improve the prediction accuracy and avoid the intrinsic defects of artificial neural network model.