计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (5): 45-50.DOI: 10.3778/j.issn.1002-8331.1507-0210

• 理论与研发 • 上一篇    下一篇

基于相关性的类偏好敏感决策树算法

周美琴,徐章艳,陈诗旭,李艳红,马  顺,展雪梅   

  1. 广西师范大学 广西多源信息挖掘与安全重点实验室,广西 桂林 541004
  • 出版日期:2017-03-01 发布日期:2017-03-03

Novel class preference sensitive decision tree algorithm based on correlation

ZHOU Meiqin, XU Zhangyan, CHEN Shixu, LI Yanhong, MA Shun, ZHAN Xuemei   

  1. Guangxi Key Lab of Multi-source Information Mining & Security, Guangxi Normal University, Guilin, Guangxi 541004, China
  • Online:2017-03-01 Published:2017-03-03

摘要: 针对决策者在面对几个分类结果时会有选择其中某一个结果的倾向性这一事实,提出了一种基于相关性的类偏好敏感决策树分类算法(CPSDT)。该算法引入了类偏好度、偏好代价矩阵等概念。为弥补在传统决策树构造过程中,选择分裂属性时未考虑非类属性之间相关性的不足,该算法在进行学习之前先采用基于相关性的特征预筛选排除属性冗余并重新构造了基于相关性的属性选择因子。经实验证明,该算法能够有效减小决策树规模,且能够在实现对偏好类的高精度预测的同时保证决策树拥有较好的整体精度。

关键词: 分类, 决策树, 属性选择因子, 偏好敏感

Abstract: In view of the fact that decision makers will have preference to one certain result when in the face of several classification results, it proposes a Class Preference Sensitive Decision Tree algorithm based on correlation(CPSDT). The algorithm introduces the concept of class-preference, the degree of class-preference and the preference cost matrix. To make up for the weakness that the correlations between non-class attributes are not considered when choosing the splitting attribute in the traditional decision tree constructing process, the algorithm uses the features pre-screened based on the correlation to exclude the redundant attributes before learning and reconstructs the attribute selection fator which is based on correlation. The experimental results show that this algorithm can reduce the size of the decision tree effectively. Further more, the algorithm can not only achieve the high precision prediction of preference class, but also can ensure the decision tree has good overall accuracy.

Key words: classification, decision tree, attribute selection factor, preference sensitive