常用费歇判别准则的比较

Comparison of several common discriminate criteria in Fisher discriminate analysis

对费歇判别分析中用于判别样品归属的几个常用判别准则进行了讨论和比较 ,认为用临界值准则判别的正确率较小 ,即使用改进的临界值准则也未必能得到较大的改善 ,而利用马氏距离法一般可得到更好的判别效果。在分析的基础上 ,提出了一种新的马氏距离法。

The article discussed and compared a few common discriminate criteria used for sample categorizing in Fisher discriminate analysis.The study showed that the result of using Mahalanobis distance criteria is better than using critical value criteria in Fisher discriminate analysis.Even if using the improved critical value criteria,the result does not improved.A new Mahalanobis distance critical is shown.