摘要
本文提出一种新的时间序列预测技术。对于一个经诊断存在混沌吸引子的时间序列,根据相空间中混沌吸引子的分形等特性,建立依赖于预测点邻界状态的预测模型;综合存在于原时间序列中确定线性趋势的外推结果,实现对原时间序列的短期预测。
This paper proposes a new method of time series forecasting, which bases on the chaotic attractor embeded in underlying time series. After having tested that there exists chaos in a detrended time series, the chaotic attractor is searched out by phase space reconstructing, and used to make short forecasting according to the nearby states of the forecasting point and the future of determinant linear trend detrended.
出处
《系统工程与电子技术》
EI
CSCD
1997年第2期23-28,共6页
Systems Engineering and Electronics
关键词
诊断
时间序列预测
预测
混沌吸引子
分形
Chaos diagnosis, Chaotic attractor, Time series, Forecasting.
