A New Weighted Hyper-Sphere Support Vector Machine

Since hyper-sphere SVM treat all samples equally, its performance is lower when distribution of the training examples is uneven. How to eliminate the influence of the uneven class sizes is important for the resulting classifier. To solve this problem, we present a new weighted hyper-sphere SVM based on the analysis of performance influence caused by the class size. Experimental results show that our method can control the misclassification rate efficiently and improve the generalization of the classifier.