An improved least squares Twin Support Vector Machine

作者：Xu Yitian^*; Xi Wenwen; Lv Xin; Guo Rui

来源：Journal of Information and Computational Science, 2012, 9(4): 1063-1071.

• Benchmark datasets
• Classification accuracy
• Classification tool
• Computational costs
• Empirical risk minimization
• Least Square
• Non-parallel planes
• Numerical experiments
• Objective functions
• Overfitting
• Quadratic programming problems
• Structural risk minimization principle
• Training errors
• Twin support vector machines
• Twin SVM

Twin Support Vector Machine (TSVM), as an effective classification tool, tries to find two non-parallel planes that can be obtained by solving two Quadratic Programming Problems (QPPs). The QPPs lead to higher computational costs. The least squares twin SVM (LS-TSVM), as a variant of TSVM, attempts to avoid the above deficiency and obtains two non-parallel planes directly by solving two sets of linear equations. However, LS-TSVM implements the empirical risk minimization principle instead of the structural risk minimization principle, i.e., only training error of samples of two classes is considered in the objective function of LS-TSVM, and it easily leads to over-fitting problem. To overcome this drawback, an improved LS-TSVM is proposed in this paper, which enhances the classification accuracy of the classifier. Numerical experiments on seven benchmark datasets demonstrate the feasibility and validity of the proposed algorithm.

• 出版日期2012
• 单位中国农业大学