一种基于单形空间缺失数据的补全方法  

Missing Data Completion Method Based on Simplex Space

作　　者：刘冰[1] 李玮琦[1] Liu Bing;Li Weiqi(Artificial Intelligence College of Dazhou Vocational and Technical College,Dazhou 635001)

机构地区：[1]达州职业技术学院人工智能学院,达州635001

出　　处：《现代计算机》2022年第19期26-32,共7页Modern Computer

基　　金：四川省教育厅自然科学重点资助项目(18ZA0217)。

摘　　要：在对基于单形空间数据进行多元回归分析时,回归系数值极不稳定,导致现有的一般统计分析方法不能恰当地对其解释和处理。本文提出了一种基于单形空间缺失数据的补全方法:首先定义了单形空间上完备的代数运算体系,然后运用K均值法对缺失数据进行初始补全,并对其进行等距对数比变换,最后运用主成分法再对初始补全值进行修正补全。实例结果表明,该补全法要比运用其他单一补全方法的效果更好。In the multiple regression analysis of data in simplex space,the value of regression coefficient is very unstable,which leads to the existing general statistical analysis methods can not properly explain and deal with it.In this paper,a method based on missing data in simplex space is proposed:firstly,a complete algebraic operation system in simplex space is defined,then the k-means method is used to complete the missing data,and the equidistant logarithmic ratio transformation is carried out.Finally,the principal component method is used to correct the initial completion value.The example results show that the effect of this method is better than that of other single methods.

关 键 词：单形空间 成分数据 对数比变换 主成分分析 数据补全 

分 类 号：O212.1[理学—概率论与数理统计]