基于改进FCM聚类的BT-SVM多类分类算法

Multi-class Classification of BT-SVM Based on Improved FCM Clustering

针对二叉树支持向量机在多类分类问题上存在的不足,利用粒子群算法对模糊C均值聚类算法进行了改进,在此基础上,结合二叉树支持向量机,构建了偏二叉树多类分类算法。该方法在二叉树各节点处根据聚类中心所对应的样本构造学习样本集和最优分类超平面,保障了聚类精度,有效地提高了测试正确率。实验表明,本文提出BT-SVM多类分类算法的测试正确率要高于同类多类分类算法。

Deflective binary tree of multi-class classification algorithm based on Binary Tree Support Vector Machine （BT-SVM） and fuzzy C-means Clustering algorithm, which overcomes defects of proposed measures effectively is put forward. The BT-SVM muticlassification is formed＇by learning sample classes and the optimal hyperplanes which are constructed in the nodes of binary tree by the samples corresponding to clustering centers. It is shown that BT-SVM muti-classification is capable of improving clustering precision by example.