类可分离性的判别，特征选择与特征提取的任务是求出一组对分类最有效的特征因此需要有定量分析比较的方法，判断所得到的特征维数及所使用特征是否对分类最有利，这种用以定量检验分类性能的准则称为类可分离性判据。

类别可分离性判据，用来检验不同的特征组合对分类性能好坏的影响，并用来导出特征选择与特征提取的方法。

理想准则：某组特征使分类器错误概率最小-Class separability of discrimination, feature selection and feature extraction task is to find a set of classification of the most effective feature thus the need for more quantitative analysis method to determine the characteristics obtained by the use of dimension and characteristics are most favorable to the classification This test for quantitative performance criteria for classification as class separability criterion. Class separability criterion, used to test different combinations of features of good and bad effects on the classification performance, and used to derive the feature selection and feature extraction methods. Ideal criteria: a set of characteristics so that the minimum classification error probability