基于树状结构的支持向量机多分类方法

Tree-based SVM Multi Classification Method

针对树状结构支持向量机多分类方法中的树结构选择问题,提出了一种基于类中心距离的树状结构选择方法,通过计算多分类问题中的各个类中心距离并对其进行排序来得到优化的树状结构,从而提高树状结构支持向量机的训练与测试效率,仿真试验证明了该方法的有效性。

Aiming at solving the permutation selection problem of the binary tree structure support vector machine, a permutation selection method was proposed based on the class center distance. Through calculating each class center distance and arranging them in a special order, the most superior tree structure will be ob- tained, thus enhanced the test accuracy of the tree structure support vector machine training and test effeicien- cy. The simulation experiment has proven the validity of the new method.