基于样本分布不平衡的近似支持向量机  

Proximal Support Vector Machines for Samples with Unbalanced Distribution

作　　者：陶晓燕[1] 姬红兵[1] 马志强[2] 

机构地区：[1]西安电子科技大学电子工程学院,西安710071 [2]空军工程大学电讯工程学院,西安710077

出　　处：《计算机科学》2007年第5期174-176,共3页Computer Science

基　　金：国防预研基金项目资助(No.51407030103DZ0117)

摘　　要：针对标准的近似支持向量机(PSVM)没有考虑样本分布不平衡的问题,提出了一种新的PSVM算法—BPS-VM;根据训练样本数量的不平衡对正负样本集分别分配了不同的惩罚因子,并将原始优化问题中的惩罚因子由数值变更为一个对角阵,最后推导出了线性和非线性BPSVM的决策函数。实验结果表明,BPSVM的性能优于PSVM,与SVM方法相比效率更高。For the problem of unbalanced data classification is not discussed in the standard Proximal Support Vector Machines, a new PSVM algorithm is presented, namely BPSVM. The different penalty factors are assigned to the positive and negative training sets according to the unbalanced population, and the penalty values are transformed into a diagonal rnatrim Finally the decision functions for the linear and nonlinear PSVM are obtained. The experimental results show that the generalization of BPSVM could be better than PSVM, and comparable to SVM with higher efficiency.

关 键 词：近似支持向量机 非平衡分布 平衡近似支持向量机 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]