Robust feature selection for multiclass Support Vector Machines using second-order cone programming

Issue title: Business Analytics in Finance and Industry January 6-8, 2014, Santiago, Chile

Guest editors: Cristián Bravo, Matt Davison, Alejandro Jofré, Sebastián Maldonado and Richard Weber

Article type: Research Article

Authors: López, Julio^a | Maldonado, Sebastián^{b; *}

Affiliations: [a] Facultad de Ingeniería, Universidad Diego Portales, Ejército 441, Santiago, Chile | [b] Universidad de los Andes, Las Condes, Santiago, Chile

Correspondence: [*] Corresponding author: Sebastián Maldonado, Universidad de los Andes, Mons. Álvaro del Portillo 12455, Las Condes, Santiago, Chile. Tel.: +56 2 26181874; E-mail:[email protected]

Abstract: This work addresses the issue of high dimensionality for linear multiclass Support Vector Machines (SVMs) using second-order cone programming (SOCP) formulations. These formulations provide a robust and efficient framework for classification, while an adequate feature selection process may improve predictive performance. We extend the ideas of SOCP-SVM from binary to multiclass classification, while a sequential backward elimination algorithm is proposed for variable selection, defining a contribution measure to determine the feature relevance. Experimental results with multiclass microarray datasets demonstrate the effectiveness of a low-dimensional data representation in terms of performance.

Keywords: Feature selection, multiclass classification, second-order cone programming, Support Vector Machines

DOI: 10.3233/IDA-150773

Journal: Intelligent Data Analysis, vol. 19, no. s1, pp. S117-S133, 2015