摘要
提出了一种改进的属性约简启发式算法 ,讨论了启发式信息的构造 .通过两个反例证明了现有的两种属性重要度定义 (基于属性依赖度的定义和基于信息熵的定义 )的不完备性 ,提出了一种加权平均的属性重要度定义 ;在此基础上构造了两种启发式算法 .通过
A modified heuristic algorithm of attribute reduction was presented. The construction of heuristic information was discussed in detail and the incompleteness of the two existing definitions of attribute significance was proved by two counterexamples. A modified definition of the attribute significance based on the weighed sum was proposed. On the basis of the definition, two heuristic algorithms were constructed. Finally, the validity and feasibility of the algorithms were demonstrated by several classical databases in the UCI repository.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2002年第4期478-481,共4页
Journal of Shanghai Jiaotong University
基金
上海市青年科学基金资助项目
关键词
粗糙集
属性约简
启发式算法
属性重要度
rough set
attribute reduction
attribute significance
heuristic algorithms
