向量相似度测度方法

Vector similarity measurement method

判断两个向量的相似程度,一般采用计算相似度的方法。首先对相似度测度法中的几种成熟的计算方法和应用优缺点进行了综合分析和对比。然后,根据实际的应用需要,结合两种成熟的距离测度法的方法,提出了一种改进的距离测度,即相对误差距离;又从放大较大相对误差作用的思路出发,提出一个相似性函数,即最大相异系数。这两种方法都充分考虑了相对变化量,具有针对性,适用于未知的待比较向量与确知的标准向量进行测度的情况。

To estimate two vectors＇ similarity, the methods of computing the similarity are generaUy adopted. Several mature similarity estimate methods and their merits and drawbacks in applications are analyzed and compared in this paper. According to the demand of practice application and associating with two mature distance measure methods, an improved distance measure method called relative error distance is proposed. And, from the thought of magnifying the effect of the larger relative error, a comparability function called most dissimilarity coefficient is created. These two methods, which have fully considered the relative variation value, are applicable in the case of measuring an unknown comparative vector and a known standard vector.