In this study we have developed a robust Support Vector Machines (SVM) scheme of classifying uncertain data. In SVM classification data uncertainty is not addressed efficiently. Furthermore, while traditional SVM methods use a single kernel for learning, multiple kernel schemes are being developed to incorporate a better understanding of all the data features. We combine the multiple kernel learning methods with the robust optimization concepts to formulate the SVM classification problem as a semi-definite programming (SDP) problem and develop its robust counterparts under bounded data uncertainties. We present some preliminary experimental results with some known datasets by introducing noise in the data. Initial analysis shows the robust SDP-SVM model improves classification accuracy for uncertain data.
CITATION STYLE
Pant, R., & Trafalis, T. B. (2015). Svm classification of uncertain data using robust multi-kernel methods. In Springer Proceedings in Mathematics and Statistics (Vol. 130, pp. 261–273). Springer New York LLC. https://doi.org/10.1007/978-3-319-18567-5_13
Mendeley helps you to discover research relevant for your work.