English

应用科学学报 ›› 1996, Vol. 14 ›› Issue (1): 48-52.

• 论文 • 上一篇    下一篇

互信息在时间序列分析中的应用

赵鸿, 柴路, 王浩, 刘书声   

  1. 天津大学
  • 收稿日期:1994-01-24 修回日期:1994-09-02 出版日期:1996-03-31 发布日期:1996-03-31
  • 作者简介:赵鸿:讲师,天津大学物理系,天津 300072
  • 基金资助:
    国家自然科学基金,攀登计划非线性科学项目资助课题

APPLICATION OF MUTUAL INFORMATION TO ANALYSING TIME SERIES

ZHAO HONG, CHAI LU, WANG HAO, LIU SHUSHENG   

  1. Tianjin University
  • Received:1994-01-24 Revised:1994-09-02 Online:1996-03-31 Published:1996-03-31

PDF

230

被引次数

15

摘要： 介绍了由信息理论定量计算两序列S、Q之间的互信息I (S,Q)=∑ijP(si,qj)·log[P(si,qj)/P(si)P(qj)]的方法;并根据周期系统、混沌系统、随机系统的一维时间序列及其时间延迟序列间的互信息估算结果,分析数据的可预测性;得出序列预测准确性的最大极限;再根据以上三类系统的二维时间序列间的互信息估算结果,估价数据所提供的信息的新颖性,从而可以避免一些重复性的工作所造成的浪费.与互相关相比,互信息能定量地提供两序列间的关系.

关键词: 互信息, 时间序列, 数据的可预测性, 数据的新颖性

Abstract: According to the information Theory, we introduce a method with which we can quantitatively calculate the mutual information of two systems S and Q,i.e.I(S; Q)=∑P(si,qj).log[P(si,qj)/P(si)P(qj)].The predictability of the data of a system was analyzed and the optimal prediction precision was obtained by the computing results of the mutual information between the one dimensional time series and its time delayed series from the time period systems,chaotic systems and random systems. The dissimilarity of the data of a system was estimated by the computing regular of the mutual information between the two dimensional time series from time period systems, chaotic systems and randoms systems,thus the waive of releasing the work could be avoided. Compared with the cross correlation function,the mutual information can quantitatively show the relation between the two time series.

Key words: time series, data predictability, data dissimilarity, mutual information