文档介绍：计算机工程 2010 年 10 月
第 36 卷第 19 期
October 2010
Engineering
· 软件技术与数据库· 文章编号:1000—3428(2010)19—0096—03 文献标识码:A 中图分类号:TP311
基于多视图树的多关系分类
郑利雄,陈琼,沈勇明
(华南理工大学计算机科学与工程学院,广州 510006)
摘要:提出一种新型的利用多视图树的方法解决多关系分类的问题。在多视图分类方法中,不同视图对分类任务贡献不同,视图间存在
互补性。对视图互补性进行研究,提出视图互补性的度量方法。根据视图互补性选择视图进行集成,构造多视图树,用于多关系分类。实
验结果表明,多视图树无论在分类效果还是效率上,都优于现有的多关系分类方法。
关键词:多关系分类;多视图;二叉树
Multi-relational Classification Based on Multiple View Tree
ZHENG Li-xiong, CHEN Qiong, SHEN Yong-ming
(School puter Science and Engineering, South China University of Technology, Guangzhou 510006, China)
【Abstract】In this paper, a new approach using multiple view tree is proposed to solve the problem of multi-relational classification. In the multiple
view classification, different individual view has different contribution to the classification task. Views plementary to the classification. The
complementary property of views is studies and pute method of views is plementary. The views are chosen to integrate to
construct a multi-view tree according to the plementary. Experimental results show that multi-relational classification method based on
multiple view tree outperforms the existing methods in terms of accuracy and efficiency.
【Key words】multi-relational classification; multiple view; binary tree
1 概述各个非目标表,各个非目标表根据目标表元组 ID 使用聚合函
随着多关系数据挖掘的兴起,各种各样的多关系分类方数便形成一个视