基于类间相似方向数的二叉树支持向量机  

Binary Tree Support Vector Machine Based on Number of Inter-class Similarity Direction

作　　者：王建建 何枫[1] 

机构地区：[1]北京科技大学东凌经济管理学院

出　　处：《统计与决策》2018年第4期15-19,共5页Statistics & Decision

基　　金：国家自然科学基金资助项目(71272160;71673022);教育部科学技术战略研究资助项目(2015KJW02);教育部科技技术委员会战略研究资助项目(KJW-A-1410;GX2015-1008(Y));中央高校基本科研业务费专项资金资助项目(FRF-BR-16-002A)

摘　　要：二叉树支持向量机是多分类的方法之一,其分类效果与二叉树结构有很大的关系。文章针对二叉树结构的分类顺序对分类精度影响的问题,改进了一种类间相似方向,并结合距离,提出了类间相似方向数作为生成偏二叉树支持向量机的多分类方法,弥补了距离不能很好地反映类分离度的缺陷。采用生成二叉树结构的方法进行数据实验分析,表明该方法具有较高的分类精度和分类效率。Binary tree support vector machine is one of the multi-classification methods, and its classification effect has much to do with the binary tree structure. Aiming at the influence of classification order of binary tree structure on classification accuracy, this paper improves a kind of inter-class similarity direction, and combines with distance to propose a multi-classification method which takes inter-class similarity direction number as generated partial binary tree support vector machine so as to overcome the defect of the distance failure to reflect the classification separation degree. Finally the paper adopts the method of generating binary tree structure to conduct a data experiment analysis, which indicates that the proposed method has higher classification accuracy and efficiency.

关 键 词：二叉树 支持向量机 类间相似方向数 多分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]