摘要
提出了利用小波变换和均生函数周期外推组合模式进行时间序列长期预测的方法.基于小波多分辨率分析理论,非平稳时间序列被分解为多个相对简单的准周期信号,信号的趋势项、周期项和随机项被分离出来.然后采用均生函数周期外推预报模式对这些准周期信号进行预报,此方法能有效的提高预报长度,并能获得较高的建模及预报精度.仿真采用两个典型实例进行验证,结果表明了方法的正确性和有效性.
A combination model forecasting approach combining wavelet transform（WT） and mean-generating function（MGF） period extrapolation is presented in the paper. According to the theory of wavelet multi-resolution analysis（MRA）, the non-stationary time series is decomposed into some relative simple and regular periodical signal series. The trend term, periodical term and stochastic term are separated from the original series. Then the mean-generating function period extrapolation forecasting mode is employed to predict these approximate periodical signals. This method can effectively improve the prediction length and has higher modeling and prediction precision. Two representative examples are adopted in the simulation experiments, the simulation results show the correctness and validity of the method.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2008年第2期283-288,共6页
Control Theory & Applications
基金
国家自然科学基金资助项目(60474014)
教育部高等学校博士学科点专项基金资助项目(20040151007)
关键词
小波变换
均生函数
周期外推
非平稳时间序列
长期预测
wavelet transform MGF period extrapolation non-stationary time series long-term forecasting
