A novel support vector machine with its features weighted by mutual information
Hong-Jie Xing, Minghu Ha, Da-Zeng Tian, Bao-Gang Hu · 2008
A novel support vector machine (SVM) with weighted features is proposed. To assign appropriate weights for each feature, a mutual information (MI) based approach is presented. Although the calculation of feature weights may add an extra computational cost, the proposed method generally exhibits better generalization performance over the traditional SVM. The numerical studies on one synthetic and five existing benchmark classification problems confirm the benefits in using the proposed method.