摘要
针对时间序列周期不等长的情况,提出了一种基于周期划分的时间序列周期分析方法.首先将时间序列变换到频域中获取序列的周期特征,其次根据周期特征计算移动平均的项数来对时间序列做移动平均处理,然后计算移动平均处理后序列中的极值点,最后对极值点按条件进行剔除后得到周期划分点.以划分点为界划分得到时间序列的多个周期段,经过分析采用周期段的中位数线来表示时间序列的周期性变化特征.这种周期划分方法更适用于存在随机波动的长序列,实验表明该方法能较好地对序列做出划分,得到的周期段中位数线的变化特点也与原时间序列基本相符.
In this paper,a method of time series period analysis based on period division is proposed for the case of unequal periods of time series.Firstly,transform the time series into frequency domain to obtain periodic characteristics of the series.Secondly,according to the periodic characteristics,the number of terms in moving average is calculated and the time series is processed by moving average.Then calculate the extreme points in the sequence after the moving average processing.Finally,dividing points are obtained after some of the extreme points are removed according to the conditions.The time series is divided into several period segments by dividing points,and the median line of the periodic segments is used to represent the periodic variation characteristics of the time series.This method is more suitable for the long series with random fluctuation.Experiments show that this method can divide the series well,and the change characteristics of the median line of the period segments obtained are basically consistent with the original time series.
作者
张稳亿
任桂山
陈建怡
李炳照
ZHANG Wen-yi;REN Gui-shan;CHEN Jian-yi;LI Bing-zhao(School of Mathematics and Statistics,Beijing Institute of Technology,Beijing 100081,China;Oil Production Technology Institute,Dagang Oilfield Company Petro China,Tianjin 300280,China)
出处
《数学的实践与认识》
2022年第6期100-109,共10页
Mathematics in Practice and Theory
基金
国家自然科学基金面上项目(62171041)
国家重点研发计划(2020YFC2006205)。
关键词
时间序列
频域分析
不等长周期
周期划分
time series
frequency domain analysis
unequal periods
period division
