摘要
在二叉树结构支持向量机(SVM)多分类算法的基础上,针对二叉树算法中点和叶盲目划分的问题,提出了一种新的二叉树SVM多分类算法。该算法通过标记划分的方法,一方面解决了盲目划分的问题,另一方面大大减少了子分类器的数目,大幅度提高了算法的速度。实验结果表明,该算法具有一定的优越性。
After studying the multi-class SVM methods based on binary tree, we propose a new multi-class SVM methods based on binary tree. This approach uses the method of tag and division and focuses on solving the problem of division the dot and the leaf blindly. It has also greatly reduced the numbers of classifier and substantially increased the speed of the algorithm. Finally, we compare the new classification algorithms with other methods. The numerical experiment results show that the multi-class SVM methods are suitable for practical use.
出处
《江南大学学报（自然科学版）》
CAS
2009年第2期145-148,共4页
Joural of Jiangnan University (Natural Science Edition)
基金
国家自然科学基金项目(60805014)
