|
计算机科学 2007
Proximal Support Vector Machines for Samples with Unbalanced Distribution
|
Abstract:
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 matrix.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.