A new algorithm of support vector machine based on weighted feature

For the classification problems based on support vector machine, if the sample contains irrelative or even completely irrelative features to the problem, the difference related to the degree of features to the problem becomes such large that may greatly affect the classification effect by means of support vector machine. To solve this problem, a new classification algorithm using SVM based on weighted features is proposed in this paper. First, the deviation between two random variables is defined, and the weights of every feature are determined by using the principle of maximizing deviations between categories, then the value to same feature for all samples is weighted by the corresponding weights of samples, respectively. Finally the samples are used for SVM training and testing. The experimental results show that the proposed algorithm can improve the classification accuracy of the classifier and decrease the numbers of support vectors.