Variable Selection in Proteomic Profile Classification by Interval Support Vector Machines (iSVM)

Abstract

For variable selection in proteomic profile classification, we present a new local modeling procedure called interval support vector machine (iSVM). This procedure builds a series of SVM models in a window that moves over the whole spectral region and then locates useful spectral intervals in terms of the least complexity of SVM models reaching a desired error level. We applied iSVM in variable selection for proteomic profile classification. The results show that the proposed procedure are very promising for classification target-based variable selection and obtain much better classification than full-spectrum SVM model.