| 基于聚类和Ripper的稀有类分类方法 |
| |
| 引用本文： | 余雯,蒋盛益,黄兴全.基于聚类和Ripper的稀有类分类方法[J].暨南大学学报,2009,30(1). |
| |
| 作者姓名： | 余雯 蒋盛益 黄兴全 |
| |
| 作者单位： | 1. 广东外语外贸大学信息学院,广东,广州,510006
2. 广东蓝鸽科技有限公司,广东,广州,510540 |
| |
| 基金项目： | 国家自然科学基金，广东省高等学校自然科学研究重点项目，广东外语外贸大学科研创新团队项目 |
| |
| 摘 要： | 稀有类分类在许多领域有重要应用,针对稀有类在数据中所占比例少,容易被忽略的特点,提出一种基于聚类和Ripper的稀有类分类方法,该方法在一趟聚类的结果中,通过将在整个数据集中所占的比例低于15%的聚类标识为少数类,再应用Ripper分类算法分别对少数类和多数类分别进行分类建模,并按照一定的组合方式调整得出整个数据集的最终规则集.在UCI数据集上的测试结果表明,基于一趟聚类和Ripper的稀有类分类方法对稀有类可产生高质量的分类效果.可以将该方法应用于现实生活的领域中进行稀有数据的分类.
|
| 关 键 词： | 数据挖掘 稀有类分类 一趟聚类 |
A rare-class classification approach based on Clustering and Ripper |
| |
| YU Wen,JIANG Sheng-yi,HUANG Xing-quan.A rare-class classification approach based on Clustering and Ripper[J].Journal of Jinan University(Natural Science & Medicine Edition),2009,30(1). |
| |
| Authors: | YU Wen JIANG Sheng-yi HUANG Xing-quan |
| |
| Institution: | 1.College of Information;Guangdong University of Foreign Studies;Guangzhou 510006;China;2.Guangdong Lancoo Co;Limited;Guangzhou 510540;China |
| |
| Abstract: | The rare-class classification is an important issue in many real life applications;this paper considers the rare-class datasets are easily ignored in the classification because of its low proportion of the whole datasets.We apply a rare-class classification approach based on clustering and Ripper.This approach is trying to find out the rare-class datasets after Cluster through recognizing every cluster whose proportion of the whole datasets is lower than 15% as the rare-class datasets.After that,Ripper algo... |
| |
| Keywords: | data mining rare-class classification One-pass Clustering |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
