互信息在时间序列分析中的应用  

APPLICATION OF MUTUAL INFORMATION TO ANALYSING TIME SERIES

作　　者：赵鸿[1] 柴路[1] 王浩[1] 刘书声[1] 

机构地区：[1]天津大学

出　　处：《应用科学学报》1996年第1期48-52,共5页Journal of Applied Sciences

基　　金：国家自然科学基金;攀登计划非线性科学项目

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．According to the information Theory, we inttoduco a methoawith whieh we ear quantitatively calculate the mutual information of two systemsS and Q,i.e.I(B; Q) =Z p(s.,qi).lug[P(B.,ql)/P(sl)p(ql)j.;The pl'edietabilityof the data of a Byetem was analyzed and the optimal prediction precision wasobtained by the combusing peEults of the mutual information between the onedimensional time Eeries and its time delayed series from the time period systems,chaotie systems and random 8}rstems. The dissimilarity of the data of a systemwas estimated by thp computing regular of the mutual information between the twodimenaional time series from time period systems, chaotic systems and randoms3tbtems,thus the waive of releasing ti.e t,rork could be avoided. Compared x-c'itlithe cross correlation function,the'mutual information can quantitatively showthe relation between the two time Eeries.

关 键 词：互信息 时间序列 数据 可预测性 信息论 

分 类 号：O236[理学—运筹学与控制论] G201[理学—数学]