摘要： 将小波分析和ARMA模型引入时间序列数据挖掘中。利用小波消噪对原始时间序列进行滤波，利用小波变换充分提取和分离金融时间序列的各种隐周期和非线性，把小波分解序列的特性和分解数据随尺度倍增而倍减的规律充分用于BP神经网络和自回归移动平均模型的建模。利用小波重构技术将各尺度域的预报结果组合成为时间序列的最终预报。经过试验验证了该方法的实际有效性。

关键词: 小波分析, ARMA模型, 神经网络, 时间序列, 数据挖掘

Abstract: This paper presents wavelet method and ARMA model for time series data mining. According to the wavelet denoising and wavelet decomposition, the hidden period and the nonstationarity existed in financial time series are extracted and separated by wavelet transformation. The characteristic of wavelet decomposition series is applied to BP networks and an Autoregressive Moving Average(ARMA) model. Finally, wavelet reconstruction is used to realize time series forcaseting. It shows that the proposed method can provide more accurate results.

Key words: wavelet analysis, Autoregressive Moving Average(ARMA) model, BP networks, time series, data mining

中图分类号: 

• TP311