摘要
属性核的确定对于求属性约简具有重要的意义,利用协调信息系统中条件属性的划分相对于决策属性划分的概率分布为单点分布的性质,提出一种新的求核算法.采用多个UCI数据集进行实验,结果表明该算法的效率高于基于辨识矩阵和基于信息熵的求核算法.
The determination of attribute core may be significantly contributory to attribute reduction. A new algorithm of computing the attribute is put forward based on the property of consistent information system that the probability distribution of condition attribution partitions relative to decision attribution partitions is one-point distribute. Several UCI data sets are used to test the approaches. The experimental and performance studies show that it is more efficient than the algorithms which based on discernibility matrix and conditional information entropy.
出处
《漳州师范学院学报(自然科学版)》
2008年第3期32-35,共4页
Journal of ZhangZhou Teachers College(Natural Science)
基金
福建省自然科学基金资助项目(2007J0039)
福建省教育厅基金资助项目(2007F5097)
关键词
粗糙集
属性核
协调信息系统
单点分布
rough set
attribute core
consistent information system
one-point distribute