摘要
提出一种基于二叉树支持向量机的超球孪生二叉树支持向量机,该算法结合了孪生支持向量机和二叉树支持向量机的优势,加快了训练速度,减少了误差累计.通过引入坐标轮换法和收缩技术,得到超球坐标轮换孪生二叉树支持向量机.实验结果表明,这两种算法具有如下优点:相比一对多支持向量机,在训练时间上具有绝对的优势,特别是在处理数据规模较大且稀疏性较强的问题时;避免了一对多支持向量机可能存在的样本不均衡性、不可分区域等缺点.
A new algorithm named as HSBT TSVM （hypersphere binary tree twin support vector machine） for multiclass classification based on BT-SVM （binary tree support vector machine） is presented in this paper, which combines the advantages of TSVM （twin support vector machine） and BT-SVM to accelerate the training speed and reduce the error accumulated. By introducing the method of coordinate rotation and shrink technology, HSCCBT-TSVM （hyper sphere cyclic coordinate binary tree twin support vector machine） is proposed to resolve the muhiclass classification problem. The experimental results show that these two algorithms have the following merits, firstly, compared to OVA-SVM （one-against-all support vector machine）, they have the absolute advantages in train data; and secondly, they avoid the imbalance property of SVM （one-against-all support vector machine）.
出处
《西南大学学报（自然科学版）》
CAS
CSCD
北大核心
2014年第7期162-168,共7页
Journal of Southwest University(Natural Science Edition)
基金
国家青年自然科学基金资助项目(11001227)
重庆市自然科学基金资助项目(CSTC2009BB2306)
中央高校基本科研业务费资助(XDJK2010B005)
关键词
孪生支持向量机
二叉树
坐标轮换法
超球体单类支持向量机
多分类
twin support vector machine
binary tree
support vector machine
multi-classification mg the time, especially for the large size and sparse sample and the indivisible region in OVA cyclic coordinate method
hypersphere one-class
