为了提高成分数据判别模型的准确性,提出一种基于等距logratio变换的成分数据判别分析方法.该建模方法先将成分数据实行等距logratio变换,在保持样本空间形态不发生变化的前提下再对变换后的数据表建立Fisher判别模型,克服了定和约束对建模的不良影响,并且保证了判别模型的合理性和准确性.应用所提出的方法,对两总体岩石分类问题建立了判别模型,并与传统模型进行比较研究.结果表明,模型的判别准确率得到很大提高,具有较强应用价值.“,”Discriminant analysis method for compositional data based on an isometric logratio transformation is proposed to improve the accuracy of compositionaldata discriminant model.This method firstly transforms the compositional data by isometric logratio transformation with no changes of sample space,and then develops Fisher discriminant model for the transformed data.It overcomes the adverse effect of the unit-sum constrain on modeling,and guarantees the rationality and accuracy of analytical results.An empirical analysis on discrimination on two types of rocks is conducted by the proposed methods and the existing method.The conclusion shows that the proposed method obtains a highly improved result and is valuable for practical application.