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,提出一种改进的 PSVM 算法(MPSVM).根据训练样本数量的不平衡对正负样本集分别分配不同的惩罚因子,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.最后推导出线性和非线性 MPSVM 的决策函数,并将其与 PSVM、非平衡的 SVM 的运算机理和性能进行比较.实验结果表明,MPSVM 的性能优于 PSVM,与非平衡 SVM 方法相比效率更高. For the standard approximate support vector machine (PSVM), the problem of unbalanced sample distribution is not considered and an improved PSVM algorithm (MPSVM) is proposed.According to the unbalance of training samples, different penalty factors are assigned to the positive and negative sample sets respectively The penalty factor in the original optimization problem is changed from a numerical value to a diagonal matrix.Finally, the decision functions of linear and nonlinear MPSVM are deduced and compared with the computational mechanism and performance of PSVM and unbalanced SVM.The experimental results show that , MPSVM outperforms PSVM and is more efficient than unbalanced SVM.