Computer Engineering and Applications ›› 2007, Vol. 43 ›› Issue (30): 242-244.

• 工程与应用 • Previous Articles     Next Articles

Prediction of time series based on Huang transform and BP neural network

CHEN Ying,XU Chen,ZHANG Wei-qiang   

  1. Institute of Intelligent Computing Science,Shenzhen University,Shenzhen,Guangdong 518060,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-10-21 Published:2007-10-21
  • Contact: CHEN Ying

基于Huang变换和BP神经网络的时间序列预测方法

陈 莺,徐 晨,张维强   

  1. 深圳大学 智能计算科学研究所,广东 深圳 518060
  • 通讯作者: 陈 莺

Abstract: Huang transform is a new method for non-stationary signal analysis developed by Norden E.Huang et al in 1998.This paper studies the application of Huang transform to time series.Firstly,the time series are decomposed into a finite and often small number of Intrinsic Mode Functions(IMF) and one Remnant Function(RF).IMF components can reflect every scaling character and RF components can represent the total trend of the origin time series.Secondly,BP neural network is applied to predict IMF and RF.Experiment results illustrate that the new predicting method is better than wavelet analysis with BP neural network and it improves the forecasting accuracy.

摘要: Huang变换是近几年发展起来处理非平稳信号的新方法。时间序列同信号一样具有非平稳的特性,研究了Huang变换在时间序列预测中的应用。首先将时间序列通过Huang变换分解为有限个固有模态函数和一个残余函数之和,每一个的固有模态函数反映了时间序列在各个尺度的特征,而残余函数则很好地反映了时间序列的总体趋势,然后应用BP神经网络对各个固有模态函数和残余函数进行预测,最后将所有的预测值重构叠加,就得到原始时间序列的预测值。实例证明,基于Huang变换和BP神经网络的时间序列的预测方法,优于小波变换和神经网络相结合的预测方法,提高了预测精度。