摘要
针对支持向量机理论中的多分类问题以及SVM对噪声数据的敏感性问题,提出了一种基于二叉树的模糊支持向量机多分类算法。该算法是在基于二叉树的支持向量机多分类算法的基础上引入模糊隶属度函数,根据每个样本数据对分类结果的不同影响,通过基于KNN的模糊隶属度的度量方法计算出相应的值,由此得到不同的惩罚值,这样在构造分类超平面时,就可以忽略对分类结果不重要的数据。通过实验证明,该算法有较好的抗干扰能力和分类效果。
An approach of binary tree fuzzy multi-class support vector machines algorithm was proposed due to the multiclassification problem and sensitivity to the noisy data of the Support Vector Machine algorithm （SVM）. The algorithm introduced a K-Nearest Neighbor （KNN） fuzzy membership function according to Binary Tree Support Vector Machine ＇algorithm （BTSVM）. Depending on the different influences of respective data set on the classification results and measuring method of the KNN fuzzy membership function, it can calculate the corresponding value and additionally obtain the different penalty value and ignore the unimportant data for the classification results, while constructing the classification hyper-planes. Experimental results indicate that the algorithm has a better ability of anti-interference and better classification effects.
出处
《计算机应用》
CSCD
北大核心
2008年第7期1681-1683,共3页
journal of Computer Applications
基金
北京市教委科技发展计划项目(KM200610028015)
国家自然科学基金资助项目(60773130)
关键词
模糊支持向量机
多分类
二叉树
fuzzy support.vector machine
multi-class
binary tree
