Proximal support vector machine for disease classification .

Venkatesan, P.; Meena Devi, G.

International Journal of Science and Technology; 2012; 2; 61-64.

Abstract: Parameter selection is one of the important steps involved in any model fitting. In this paper we have used Uniform Design Tables to choose the parameters for PSVM and SVM to classify the data. UD is one of the efficient space filling designs, which spreads the combination of parameters in the space uniformly scattered and generalizes the performance of the model efficiently. This paper compares performance of Proximal SVM and SVM with respect to prediction accuracy, using clinical trial outcomes on spinal tuberculosis treated patients. The PSVM performs slightly better than SVM in the training set, but the performance is as good as SVM in the test set.

Keywords: Uniform design tables; SVM; Proximal SVM (PSVM)

Back to List of publications / Home