中文摘要：

为克服传统支持向量机不能处理交叉数据分类问题，Mangasarian等人提出一种新的分类方法PSVM，该方法可有效解决交叉数据两分类问题，但用PSVM解决多分类问题还报道不多。为此，提出一种基于PSVM的多分类方法（M—PSVM），并探讨训练样本比例与分类精度之间关系。在UCI数据集上的测试结果表明，M—PSVM与传统SVM分类性能相当，且当训练样本比例小时，效果更优；此外，在入侵检测数据集上的初步实验表明，M—PSVM可有效改进少数类的分类精度，因而为求解数据不平衡下的分类问题提供了新的思路，进一步的实验验证正在进行。

英文摘要：

Mangasarian and Wild proposed a new classification method PSVM to handle the classification of cross-data that the traditional support vector machine couldn＇t overcome,but very litter work has been done for the multi-classification using PSVM. In this paper,based on PSVM,we propose a new multi-classification method（M-PSVM）,and discuss the relationship between the rate of train data and the accuracy.Experimental results on UCI show that,the classification effect of M-PSVM method is as well as C-SVM,sometimes the smaller training data is,the better performance M-PSVM has;Experimental results in the intrusion detection datasets show that,M-PSVM can effectively improve the classification accuracy of the very small classes,hence which provides a new strategy for solving unbalanced data classification,further experimental testing is ongoing.