计算机科学与探索 ›› 2014, Vol. 8 ›› Issue (9): 1137-1145.DOI: 10.3778/j.issn.1673-9418.1406013

• 人工智能与模式识别 • 上一篇    下一篇

名词性数据的五种代价敏感属性约简算法比较

林姿琼+,李静宽,赵  红   

  1. 闽南师范大学 粒计算及其应用重点实验室,福建 漳州 363000
  • 出版日期:2014-09-01 发布日期:2014-09-03

Comparison of Five Cost-Sensitive Attribute Reduction Algorithms for Nominal Data

LIN Ziqiong+, LI Jingkuan, ZHAO Hong   

  1. Lab of Granular Computing, Minnan Normal University, Zhangzhou, Fujian 363000, China
  • Online:2014-09-01 Published:2014-09-03

摘要: 代价敏感学习是数据挖掘研究领域最具有挑战性的问题之一。属性约简是数据挖掘中重要的经典问题。代价敏感属性约简问题是对经典属性约简问题的自然扩展,已经逐渐成为研究的热点。对当前具有代表性的5种处理名词性数据的代价敏感属性约简算法进行了分析和比较,总结了每种算法的各方面特性及不足之处,从而便于研究者对已有算法进行改进,并且进一步提出具有更好性能的新的约简算法,方便用户对算法的选择和使用。

关键词: 数据挖掘, 代价敏感, 属性约简, 最优因子

Abstract: Cost-sensitive learning is one of the most challenging problems in the current stage of data mining and machine learning research. Attribute reduction is an important classical problem of data mining. Cost-sensitive attribute reduction problem is an extension of traditional attribute reduction problem, and has become a hot research object. This paper analyzes and compares current typical five cost-sensitive attribute reduction algorithms for nominal data, and summarizes the main features and shortcomings of five algorithms, thereby researchers can improve old algorithms or develop new effective ones. The summary can also be used to select data mining techniques for new applications.

Key words: data mining, cost sensitive, attribute reduction, optimal factor