A New Weighted Nu-Support Vector Machine

v-Support vector machine(v-SVM) is one of the most widely used support vector machines for providing a way to control the classification precision of training. However, its different classification precision of the two training classes with different sizes and neglect of the important level of training samples prevent it from being applied to some applications in which the training classes´ sizes are uneven and the importance of samples is different from each other. In this paper, a class and sample weighted v-SVM is proposed. It introduces class weights and sample weights into the error penalty part of the objective function. Theoretical analysis shows that class weights can be used to control the classification precision of each class, and sample weights can be used to increase the probability of correct classification of some samples.