基于随机下采样和SMOTE的不均衡SVM分类算法

摘要

传统的支持向量机( SVM)算法在数据不均衡的情况下,分类效果很不理想.为了提高SVM算法在不均衡数据集下的分类性能,提出随机下采样与SMOTE算法结合的不均衡分类方法.该方法首先利用随机下采样对多数类样本进行采样,去除样本中大量重叠的冗余样本,使得在减少数据的同时保留更多有用信息；而对少数类样本则是利用SMOTE算法进行过采样.实验部分将其应用在UCI数据集中并同其他采样算法比较,结果表明文中算法不但能有效提高SVM算法在不均衡数据中少数类的分类性能,而且总体分类性能也有所提高.%The classification result of classical support vector machine ( SVM) algorithm in the case of unbalanced data set is not satisfactory. In order to improve the SVM algorithm ' s classification performance under unbalanced data set, a novel algorithm based on random under-sampling and synthetic minority over-sample technique ( SMOTE ) is presented in this paper. The Random under-sampling is firstly applied to under-sample the majority class instances for removal of a large number of overlapping samples of redundant and noise samples, consequently making reservations for the majority class instances with more useful information; and then SMOTE is used to over-sample the minority class boundary instances, such that can be more conducive to generate the effective classification interface of SVM algorithm. The experimental results on UCI datasets compared with other sampling algorithms show that the proposed method can not only improve classification performance of SVM in the minority class data, but also increase the overall classification performance.