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

Research areas:
  • Uncategorized
Year:
2015
Type of Publication:
Article
Authors:
  • J López
  • S. Maldonado
Journal:
Intelligent DataAnalysis
Volume:
19
Pages:
S117-S133
Abstract:
This work addresses the issue of high dimensionality for linear multiclass Support Vector Machines (SVMs) using second-order cone programming (SOCP)formulations. Theseformulationsprovide arobustand efficientframework for classification, whilean adequate feature selection process may improve predictive performance. Weextend theideas 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 alow-dimensional data representation interms of performance.