Least squares support vector machine with parametric margin for binary classification¹

Article type: Research Article

Authors: Yang, Zhixia^{a; b; *} | Zhou, Zhe^a | Jiang, Yaolin^a

Affiliations: [a] College of Mathematics and Systems Science, Xinjiang University, Urumuqi, P.R.China | [b] State Key Lab of Biochemical Engineering, Institute of Process Engineering, Chinese Academy of Science, Beijing, China

Correspondence: [*] Corresponding author. Zhixia Yang, E-mail: xjyangzhx@sina.com.

Note: [1] This work is supported by the National Natural Science Foundation of China (No. 11161045), the China Postdoctoral Science Foundation (No. 2015M572625), the Project-sponsored by SRF for ROCS, SEM and Open Funding Project of the National Key Laboratory of Biochemical Engineering (No. 2013KF-01)

Abstract: In this paper, we propose a least squares support vector machine with parametric margin (Par-LSSVM) for binary classification, which only needs to solve a system of linear equation. Par-LSSVM is able to handle the datasets with heteroscedastic noise. And the closer hyperplane to the test data point gives the class label, and this makes Par-LSSVM capable of dealing with “Cross Planes” datasets. The experimental results on several artificial, benchmark and USPS datasets indicate that our proposed algorithm outperforms Par-ν-SVM for binary classification problem.

Keywords: Support vector machine, classification, least squares, parametric margin

DOI: 10.3233/IFS-151743

Journal: Journal of Intelligent & Fuzzy Systems, vol. 30, no. 5, pp. 2897-2904, 2016