小波包与神经网络相结合的股票价格预测模型

小波包较之于小波可以更为灵活地提取分散在不同尺度上的信号特征，结合神经网络也就可获得更好的预测精度，本文按此方式建立了一种混合杂交模型用于股票市场价格波动预测，并为获得最优预测精度，本文利用遗传算法进行小波包最优分解选择和神经网络参数选择。通过对上证综指的实证研究，表明这种混合杂交模型的性能优于同类神经网络模型和基于小波分解的神经网络模型。

Forecasting Model of Stock Price by Wavelet Packet Integrated Neural Network

Compared with the wavelet's fix decomposing method, wavelet packet is a more flexible decomposing method to extract the features in the multiscales. Therefore wavelet packet integrated neural network will get more precise forecasting result. Based on it a model for forecasting stock price is built up in this paper. To get the optimal forecasting accuracy, GA is also used in the selecting wavelet packet decomposing and neural network parameters. Case study in the SZZI index shows that this hybrid model is more competitive than other similar neural network model and wavelet integrated neural network model.