¹School of Computer Science and Technology, China University of Mining and Technology, Xuzhou, China
²Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, 100190 China

---

Abstract—Twin Support Vector Machines (TWSVM) are developed on the basis of Proximal Support Vector Machines (PSVM) and Proximal Support Vector Machine based on the generalized eigenvalues(GEPSVM). The solving of binary classification problem is converted to the solving of two smaller quadratic programming problems by TWSVM. And then it gets two non-parallel hyperplanes. Its efficiency of dealing with the problems and performance are better than the traditional support vector machines. However, it also has some problems. Its own parameters are difficult to be appointed. In order to solve this problem, on the basis of in-depth study of TWSVM, this paper proposes an algorithm that is the Twin Support Vector Machines based on Quantum Particle Swarm Optimization (QPSOTWSVM) .By the use of the global searching ability of the Quantum Particle Swarm Optimization (QPSO), QPSOTWSVM can search the optimal parameters in the global scope and avoid itself falling into the local optimum prematurely to find the values of the parameters which are the closest to the optimal parameters. QPSO-TWSVM avoids using the empirical values to appoint the parameters successfully. Compared with the traditional TWSVM, QPSO-TWSVM can appoint the parameters more accurately and avoid selecting the parameters blindly. Because of the better parameter selections, QPSO-TWSVM improves the classification accuracy of TWSVM.

Index Terms—QPSO, TWSVM, parameter optimization, binary classification.

[PDF]

Cite: Shifei Ding, Fulin Wu, Ru Nie, Junzhao Yu, Huajuan Huang, "Twin Support Vector Machines Based on Quantum Particle Swarm Optimization," Journal of Software vol. 8, no. 7, pp. 1743-1750, 2013.