Feature selection for linear SVMs under uncertain data: Robust optimization based on difference of convex functions algorithms

• Published in: Neural networks Vol. 59; pp. 36 - 50
• Main Authors: Le Thi, Hoai An, Vo, Xuan Thanh, Pham Dinh, Tao
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.11.2014; Elsevier
• Subjects: Algorithmics. Computability. Computer arithmetics; Algorithms; Applied sciences; Computer Science; Computer science; control theory; systems; Data Interpretation, Statistical; Data processing. List processing. Character string processing; DC programming; DCA; Decision theory. Utility theory; Exact sciences and technology; Feature selection; Humans; Leukemia - genetics; Linear Models; Memory organisation. Data processing; Microarray Analysis; Operational research and scientific management; Operational research. Management science; Reliability theory. Replacement problems; Robust optimization; Software; Support Vector Machine; SVM; Theoretical computing; Weather; DC programming; Feature selection; Robust optimization; DCA; SVM; Difference of convex functions; Modeling; Selection problem; Uncertain system; Vector support machine; Selection criterion; Robustness; Data models; Ellipsoid; Mathematical programming
• ISSN: 0893-6080; 1879-2782; 1879-2782
• DOI: 10.1016/j.neunet.2014.06.011

In this paper, we consider the problem of feature selection for linear SVMs on uncertain data that is inherently prevalent in almost all datasets. Using principles of Robust Optimization, we propose robust schemes to handle data with ellipsoidal model and box model of uncertainty. The difficulty in treating ℓ0-norm in feature selection problem is overcome by using appropriate approximations and Difference of Convex functions (DC) programming and DC Algorithms (DCA). The computational results show that the proposed robust optimization approaches are superior than a traditional approach in immunizing perturbation of the data.