近似周期时间序列的周期识别及提取

发布时间：2018-04-13 13:14

本文选题：近似周期时间序列 + 周期识别及提取　；参考：《华东师范大学》2015年硕士论文

【摘要】：对时间序列周期性的研究是现下研究的一个热点.目前生活中的许多事件和现象,它们的历史数据形成的时间序列存在一定的周期特征,但是有很多周期特征表现出周期长度不相等的现象.基于这方面的考虑,本文主要研究周期长度不固定的时间序列的周期性,即近似周期时间序列的周期识别及提取.本文首先介绍了近似周期时间序列的概念,对已被提出的基于矩估计方法提取时间变换函数做了简单的介绍.并在此基础上,提出了一种新的方法估计时间变换函数——拟合估计方法.对于拟合估计方法,拟合数据的提取非常重要,因此本文在第三章给出了两种拟合数据的选取方法,并证明了第二种方法选取出的数据能够真实的反映时间变换.最后将两种估计时间变换函数的方法通过实证分析进行比较,说明了基于拟合估计方法得到的结果优于基于矩估计方法得到的结果.本文最后指出一个现象,不同取样方式对时间序列周期识别是有影响的.指出对于某些取样方式,得到的时间序列的周期无法反映其真实周期.另外,由于时间序列噪声的存在,本文给出的拟合数据选取方法在某些程度上会受到影响,还需要不断改进.
[Abstract]:The research on the periodicity of time series is a hot topic.At present, many events and phenomena in life, the time series formed by their historical data have some periodic characteristics, but many of the periodic characteristics show the phenomenon that the period length is not equal.Based on these considerations, this paper mainly studies the periodicity of time series with unfixed cycle length, that is, the periodic identification and extraction of approximate periodic time series.In this paper, the concept of approximate periodic time series is introduced, and the time transform function extraction based on moment estimation is briefly introduced.On the basis of this, a new method for estimating time transform function-fitting estimation is proposed.For fitting estimation, the extraction of fitting data is very important, so this paper gives two methods of selecting fitting data in the third chapter, and proves that the data selected by the second method can truly reflect the time transformation.Finally, two methods of estimating time transform function are compared by empirical analysis, and the results based on fitting estimation method are better than those based on moment estimation method.Finally, this paper points out a phenomenon that different sampling methods have influence on time series period identification.It is pointed out that the period of time series can not reflect the true period for some sampling methods.In addition, due to the existence of time series noise, the fitting data selection method presented in this paper will be affected to some extent and needs to be improved.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O211.61

【参考文献】