支持向量机在处理非平衡数据集时常常不能取得良好的效果,因为其分类性能只考虑了总体分类精度,而忽略了不同类别样例之间的精度权衡.本文提出了一种基于样例分布的样例惩罚支持向量机,可以针对每一个样例根据其相应的分布特性选取惩罚以获得高敏感度的分类面.实验表明,该模型比标准支持向量机在非平衡数据上具有更好的性能. SVM often can not get good results when dealing with unbalanced datasets because its classification performance only considers the overall classification accuracy and neglects the precision trade-off between different types of samples.This paper presents a sample distribution , We can choose the penalty for each sample according to its corresponding distribution characteristics to obtain a high sensitivity classification surface.Experiments show that this model has better performance than the standard support vector machine on unbalanced data performance.