Using a support vector machine and sampling to classify compounds as potential transdermal enhancers

Abstract

Distinguishing good chemical enhancers of percutaneous absorption from poor enhancers is a difficult problem. Previously, discriminant analysis and other machine learning methods have been applied to this problem. Results showed that the ordinary SVM provided the best result. In this work, we apply both SVM with different cost errors and sampling methods to improve the accuracy of classification. We show that a good classification is possible. © 2012 Springer-Verlag.