Fuzzy chance constrained least squares twin support vector machine for uncertain classification¹

Issue title: Applied Mathematics Related to Nonlinear Problems

Guest editors: Juan L.G. Guirao and Wei Gao

Article type: Research Article

Authors: Han, Renjie^a | Cao, Qilin^{b; *}

Affiliations: [a] College of Economics, Sichuan University, Chengdu, China | [b] Business School, Sichuan University, Chengdu, China

Correspondence: [*] Corresponding author. Qilin Cao, Business School, Sichuan University, Chengdu, China. E-mail: qlcao@scu.edu.cn.

Note: [1] This work was supported by the Fundamental Research Funds for the Central Universities of Sichuan University (skqy201646, skgt201602).

Abstract: In this paper, via chance constrained programming formulation and fuzzy membership, we give suggestions on a new fuzzy chance constrained least squares twin support vector machine, which can make data measurement noise efficiently. In this paper, we concentrate on least squares twin support vector machine classification when data distributions are uncertain statistically. The model’s function is used to guarantee the small probability of misclassification for the uncertain data, with some known characters of the distribution. The fuzzy chance constrained least squares twin support vector machine model can be transformed into second-order cone programming (SOCP) through the properties of moment information of uncertain data and thus the dual problem of SOCP model is introduced. Besides, through the numerical experiments we also demonstrate the model’s performance in real data and artificial data.

Keywords: Support vector machine, robust optimization, chance constraints, uncertain classification

DOI: 10.3233/JIFS-169355

Journal: Journal of Intelligent & Fuzzy Systems, vol. 33, no. 5, pp. 3041-3049, 2017