Twin SVM with a reject option through ROC curve

• Main Authors: Lin, Dongyun, Sun, Lei, Toh, Kar-Ann, Zhang, Jing Bo, Lin, Zhiping
• Other Authors: School of Electrical and Electronic Engineering
• Format: Article
• Language: English
• Published: 2018
• Subjects: ROC Curves; Twin SVM
• Online Access: https://hdl.handle.net/10356/86871; http://hdl.handle.net/10220/44245
• Institution: Nanyang Technological University

This paper proposes a new method which embeds a reject option in twin support vector machine (RO-TWSVM) through the Receiver Operating Characteristic (ROC) curve for binary classification. The proposed RO-TWSVM enhances the classification robustness through inclusion of an effective rejection rule for potentially misclassified samples. The method is formulated based on a cost-sensitive framework which follows the principle of minimization of the expected cost of classification. Extensive experiments are conducted on synthetic and real-world data sets to compare the proposed RO-TWSVM with the original TWSVM without a reject option (TWSVM-without-RO) and the existing SVM with a reject option (RO-SVM). The experimental results demonstrate that our RO-TWSVM significantly outperforms TWSVM-without-RO, and in general, performs better than RO-SVM.