Application research on semi-definite programming optimized support vector machines

ABSTRACT

The kernel function optimization is the key issues to address when using the support vector machine (SVM) algorithm. To solve the parameter selection for the SVM, a semi-definite programming optimized SVM (SDP-SVM) algorithm is proposed in this paper. The steps of the algorithm are described, and the optimization of the kernel function is shown using an SDP method. The SDP method is used to find the best parameter of SVM. The heart_scale data in the University of California Irvine database are then simulated using the SDP-SVM model. The experimental results shows that the generalization capability and the classification accuracy of the SDP-SVM algorithm have been greatly improved. A variety of strip-steel surface defect images from actual production are classified using the SDP-SVM algorithm, and the results show that the classification method of the SDP-SVM algorithm has high classification accuracy, strong practicability, and a wide variety of application prospects.

KEYWORDS:

• Support vector machines
• semi-definite programming
• kernel function
• strip-steel defect image

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Jingzhong Hou http://orcid.org/0000-0003-2226-7438

Notes on contributors

Jingzhong Hou is a PhD student in School of Electronics and Information Engineering, Hebei University of Technology, Tianjin, China. His current research interests include machine learning.

Kewen Xia received the PhD degree in electronic science and technology from Xi'an Jiaotong University, China in 2003. Now he works as a professor and supervisor of PhD candidates at Hebei University of Technology, China. His current research interests cover intelligent information processing and communication technology.

Fan Yang received PhD degree in Electrical Engineering and Automation from Harbin Institute of Technology, China, in 2004. Now he works as a professor and supervisor of PhD candidates at Hebei University of Technology, China. His current research interests include intelligent information processing and machine learning.

Additional information

Funding

This work was supported by the Natural Science Foundation of Hebei Province (CN) (E2016202341) and the Research Project of Science and Technology for Hebei Province Higher Education Institutions (BJ2014013).