计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (14): 20-23.DOI: 10.3778/j.issn.1002-8331.2010.14.006

• 博士论坛 • 上一篇    下一篇

基于覆盖的多标记学习方法研究

段 震1,2,程家兴1,2,张 铃1,2   

  1. 1.安徽大学 计算智能与信号处理教育部重点实验室,合肥 230039
    2.安徽大学 计算机科学与技术学院,合肥 230039
  • 收稿日期:2010-01-29 修回日期:2010-03-12 出版日期:2010-05-11 发布日期:2010-05-11
  • 通讯作者: 段 震

Research of multi-label learning based on covering algorithm

DUAN Zhen1,2,CHENG Jia-xing1,2,ZHANG Ling1,2   

  1. 1.Key Lab of Intelligent Computing & Signal Proceesing,Ministry of Education,Anhui University,Hefei 230039,China
    2.School of Computer Science and Engineering,Anhui University,Hefei 230039,China
  • Received:2010-01-29 Revised:2010-03-12 Online:2010-05-11 Published:2010-05-11
  • Contact: DUAN Zhen

摘要: 多标记学习是实际应用中的一类常见问题,覆盖算法在单标记学习中表现出了优秀的性能,但无法处理多标记情况。将覆盖算法推广到多标记学习中,针对多标记学习的特点和评价指标,对算法的学习和构造过程进行了改造,给出待分类样本对各类别的隶属度。将算法应用于基因数据集和自然场景数据集的学习中,实验结果表明算法能够取得较好的分类效果,且相比于大多数同类算法有更高的性能。

关键词: 覆盖算法, 核函数, 多标记, 基因分类

Abstract: Multi-label learning is a common problem in real application.Covering algorithm performs well with single-label learning but can not deal with multi-label learning.In this paper,covering algorithm is extended to Multi-label Learning Covering Algorithm(MLCA).Training and testing procedures are adapted to the characteristics of multi-label learning problem,and the membership function of sample is calculated.MLCA is applied to the gene classification and nature scene classification and the results show that MLCA is effective and has better performance than many other learning algorithms in the field.

Key words: covering algorithm, kernel function, multi-label, gene classification

中图分类号: