Comparison of performance with support vector machines

ABSTRACT Support vector machine (SVM) is an important machine learning method, which has many applications in pattern recognition, network security, etc. However, this method has some shortcomings such as complicated computation of quadratic programming, time-consuming training and low anti-noise performance. To this end, the researchers have proposed some improved methods. In this paper, we select the commonly used SVM classifiers including C-SVM, v-SVM, PSVM and TWSVM to study their performance of classification on the standard data set by experiments. The experimental results are shown that the performance with TWSVM has an advantage over C-SVM, v-SVM and PSVM in selected ten data sets using linear kernel. However, when Gauss kernel is used, accuracy with different support vector machine is almost no differences except data set wdbc.