摘要
提出了一种基于经验模态分解(Empirical Mode Decomposition,EMD)技术的时间序列组合预测模型。首先对非平稳非线性时间序列进行EMD技术分解,然后将分解得到的子序列进行聚类,并运用传统的时间序列预测方法对各子序列分别进行预测,最后汇总子序列的预测值得到目标时间序列的预测值。统计模拟和实证分析显示:组合预测模型能够显著提高预测的精度,说明新方法对于非平稳非线性时间序列的预测是有效的。
This paper presents a combined forecasting model of time series based on EMD technology.Firstly,we decompose non-linear and non-stationary time series using EMD.Then we cluster the sub-sequences and use traditional time series prediction methods to predict each sub-sequence.Finally,we predict the target time series by adding up the predictions of sub-sequences.Statistical simulation and empirical analysis validate that the combined forecasting model can improve the prediction accuracy,thus the new method is effective for predicting non-linear and non-stationary time series.
出处
《系统工程》
CSSCI
CSCD
北大核心
2014年第5期138-143,共6页
Systems Engineering
基金
国家社会科学基金重大项目(13&ZD148)
关键词
经验模态分解
非平稳非线性时间序列
人工智能法
移动平均法
Empirical Mode Decomposition
Nonlinear and Non-stationary Time Series
Artificial Intelligence
Moving Average Method