A Novel Evaluation Method Basing on Support Vector Machines

Abstract

Recently support vector machine (SVM) has become a more and more popular classification tool. We presented our two-phase, efficient, and fair evaluation method for DRMs (digital right management system) basing on SVM. Influence of three difference methods and test set number on evaluation result is discussed. After analysized by binary logistic regression, odds ratio comparison of SVM with multi-phase fuzzy synthesized evaluation and FCM illustrates that SVM is the most excellent in these three approaches. Through detailed experimental evaluations under various data set of samples and approaches, our evaluation method of SVM is illustrated to be scalable and accurate.