基于特征加权的多核支持向量机  

A multi-core support vector machine based on feature weighted

作　　者：邵朝[1] 李强[1] 

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710121

出　　处：《西安邮电大学学报》2017年第2期84-88,共5页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金重点资助项目(61136002);陕西省自然科学基金资助项目(2014JM8331;2014JQ5183;2014JM8307);陕西省教育厅科学研究计划资助项目(2015JK1654)

摘　　要：提出一种改进的多核支持向量机算法。利用特征向量的信息熵,计算出特征向量关于分类标签的信息增益,用该信息增益作为数据集特征向量的权重系数,然后结合多核支持向量机的核学习能力,对特征向量非均等区别对待,使得分类更加准确。对比实验结果表明,与支持向量机、多核支持向量机相比,改进算法分类正确率得以提升。An improved multi kernel support vector machine algorithm is proposed in this paper. Based on the information entropy feature vector, the information gain feature vector on the classification label is calculated. Using the information gain as the weight coefficient data set of feature vectors and then combined with the multi kernel support vector machine kernel learning ability, the feature vector of non equal distinction is obtained to make the class more accurate. Experimental results show that the classification accuracy of the improved algorithm is improved compared with the support vector machine （SVM） and the multi kernel support vector machine （SVM）.

关 键 词：特征加权 多核 支持向量机 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]