基于二叉树的多分类SVM算法在电子邮件过滤中的应用

目前性能较好的多分类算法有1-v-r支持向量机（SVM）、1-1-1SVM、DDAG SVM等，但存在大量不可分区域且训练时间较长的问题。提出一种基于二叉树的多分类SVM算法用于电子邮件的分类与过滤，通过构建二叉树将多分类转化为二值分类，算法采用先聚类再分类的思想，计算测试样本与子类中心的最大相似度和子类间的分离度，以构造决策节点的最优分类超平面。对于C类分类只需C-1个决策函数，从而可节省训练时间。实验表明，该算法得到了较高的查全率、查准率。

Application of multiclass SVM based on binary tree in E-mail filtering

Now some preferable performance multiclass algorithms, such as 1-v-r support vector machine (SVM) , 1-1-1 SVM and DDAG SVM，have many problems of impartible regions and longer training time.A new multiclass SVM algorithm based on binary tree was introduced on E-mail filtering. It could convert multiclass problem to binary classification by constructing binary tree. The idea of clustering first and classification later was adopted, and the largest similarity between testing sample and sub-category center and the separation measure of sub categories were calculated, in oder to construct the optimal class hyperplane of decision making nodes. Only C-1 optimal functions were needed for C kinds of classification, so training time could be saved. The experiment results show that the new algorithm has higher filtering recall and precision.