AN INTEGRATED DATA CHARACTERISTIC TESTING SCHEME FOR COMPLEX TIME SERIES DATA EXPLORATION
AN INTEGRATED DATA CHARACTERISTIC TESTING SCHEME FOR COMPLEX TIME SERIES DATA EXPLORATION
复制标题
复杂时间序列数据探索的综合数据特征测试方案
DOI：
10.1142/s0219622013500193
复制
发表时间：
2013-05-01
影响因子：
4.9
通讯作者：
Xu, Weixuan
中科院分区：
文献类型：
--
作者：
Tang, Ling;Yu, Lean;Xu, Weixuan
为了选择最适合复杂时间序列建模的研究方法，提出了一种用于复杂时间序列数据探索的综合数据特征测试方案。根据不同数据特征之间的关系，本文将时间序列数据特征分为两大类：性质特征和模式特征。因此，在所提出的测试方案中涉及两个相关的任务：性质确定和模式度量。在定性时，对产生时间序列数据的动力学系统进行非平稳性、非线性和复杂性检验。在模式度量中，根据模式重要性来衡量周期性(和季节性)、易变性(或突变)和随机性(或噪声模式)的特征。为了说明这一点，本文以四个主要的中国经济时间序列数据为检验对象，利用所提出的综合检验方案，深入挖掘了这些时间序列数据中所隐藏的数据特征。实证结果表明，在定性阶段，所有样本数据的性质都表现出了复杂性，同时基于模式重要度捕捉了每个时间序列的主要模式，表明该方法可以作为一种有效的数据特征测试工具，用于复杂时间序列数据的综合探索。
In this paper, an integrated data characteristic testing scheme is proposed for complex time series data exploration so as to select the most appropriate research methodology for complex time series modeling. Based on relationships across different data characteristics, data characteristics of time series data are divided into two main categories: nature characteristics and pattern characteristics in this paper. Accordingly, two relevant tasks, nature determination and pattern measurement, are involved in the proposed testing scheme. In nature determination, dynamics system generating the time series data is analyzed via nonstationarity, nonlinearity and complexity tests. In pattern measurement, the characteristics of cyclicity (and seasonality), mutability (or saltation) and randomicity (or noise pattern) are measured in terms of pattern importance. For illustration purpose, four main Chinese economic time series data are used as testing targets, and the data characteristics hidden in these time series data are thoroughly explored by using the proposed integrated testing scheme. Empirical results reveal that the natures of all sample data demonstrate complexity in the phase of nature determination, and in the meantime the main pattern of each time series is captured based on the pattern importance, indicating that the proposed scheme can be used as an effective data characteristic testing tool for complex time series data exploration from a comprehensive perspective.