Support Vector Machines Based on Reducing Noise

Support vector machine (SVM) is one of the standard tools for machine learning. Based on the statistical learning theory and optimization theory. But, SVM, which still has some limitations, Support vector machine is the algorithm to solve two classification problems, and for multi-class classification problems, there are many work to need further research. Especially Support vector machines is very sensitive to noises and outliers, how to reduce the effect of nises and outliers needs to be further explored. Support Vector Machine (SVM) is sensitive to noises and outliers. For reducing the effect of noises and outliers, we propose a novel SVM based on reducing noises and outliers. We propose a new objective function and control the size of error function. The importance and error of samples are considered. By solving the dual problem of SVM, the separation hypersurface is simplified and the margin of hypersurface is widened. The numbers of support vector and running times are decreased. Experimental results show that our proposed method is able to simultaneously increase the classification efficiency and the generalization ability of the SVM.