中国物理B ›› 2009, Vol. 18 ›› Issue (7): 2690-2695.doi: 10.1088/1674-1056/18/7/011

• • 上一篇    下一篇

Fine-grained permutation entropy as a measure of natural complexity for time series

王越1, 刘小峰2   

  1. (1)The Key Laboratory of Robot and Intelligent Technology of Shandong Province, and College of Information and Electrical Engineering, Shandong University of Science and Technology, Qingdao 266510, China; (2)The Key Laboratory of Robot and Intelligent Technology of Shandong Province, and College of Information and Electrical Engineering, Shandong University of Science and Technology, Qingdao 266510, China;Institute of Artificial Intelligence and Robot, Xi'an Jiaotong University, Xi'an 710049, China
  • 收稿日期:2008-11-10 修回日期:2009-01-11 出版日期:2009-07-20 发布日期:2009-07-20
  • 基金资助:
    Project supported by the National High Technology Research and Development Program of China (Grant No 2007AA04Z238) and the Qingdao Foundation for Development of Science and Technology, China (Grant No 06-2-2-10-JCH).

Fine-grained permutation entropy as a measure of natural complexity for time series

Liu Xiao-Feng(刘小峰)a)b)† and Wang Yue(王越)a)   

  1. a The Key Laboratory of Robot and Intelligent Technology of Shandong Province, and College of Information and Electrical Engineering, Shandong University of Science and Technology, Qingdao 266510, China; b Institute of Artificial Intelligence and Robot, Xi'an Jiaotong University, Xi'an 710049, China
  • Received:2008-11-10 Revised:2009-01-11 Online:2009-07-20 Published:2009-07-20
  • Supported by:
    Project supported by the National High Technology Research and Development Program of China (Grant No 2007AA04Z238) and the Qingdao Foundation for Development of Science and Technology, China (Grant No 06-2-2-10-JCH).

PDF (PC)

1151

赞

摘要： In a recent paper [2002 Phys. Rev. Lett. 88 174102], Bandt and Pompe propose permutation entropy (PE) as a natural complexity measure for arbitrary time series which may be stationary or nonstationary, deterministic or stochastic. Their method is based on a comparison of neighbouring values. This paper further develops PE, and proposes the concept of fine-grained PE (FGPE) defined by the order pattern and magnitude of the difference between neighbouring values. This measure excludes the case where vectors with a distinct appearance are mistakenly mapped onto the same permutation type, and consequently FGPE becomes more sensitive to the dynamical change of time series than does PE, according to our simulation and experimental results.

Abstract: In a recent paper [2002 Phys. Rev. Lett. 88 174102], Bandt and Pompe propose permutation entropy (PE) as a natural complexity measure for arbitrary time series which may be stationary or nonstationary, deterministic or stochastic. Their method is based on a comparison of neighbouring values. This paper further develops PE, and proposes the concept of fine-grained PE (FGPE) defined by the order pattern and magnitude of the difference between neighbouring values. This measure excludes the case where vectors with a distinct appearance are mistakenly mapped onto the same permutation type, and consequently FGPE becomes more sensitive to the dynamical change of time series than does PE, according to our simulation and experimental results.

Key words: complexity, entropy, dynamical change, fine-grained symbolization

中图分类号:  (Time series analysis)

  • 05.45.Tp
02.50.Ey (Stochastic processes) 05.70.Ce (Thermodynamic functions and equations of state)