基于知识粒度的不完备决策表的属性约简的矩阵算法  

Discernibility Matrix Algorithm of Attribute Reduction Based on Knowledge Granulaion in Incomplete Decision Table

作　　者：张清国[1] 郑雪峰[1] 

机构地区：[1]北京科技大学信息工程学院,北京100083

出　　处：《计算机科学》2012年第2期209-211,243,共4页Computer Science

基　　金：国家科技基础条件平台项目(2005DKA43600);国家自然科学基金项目(60674054)资助

摘　　要：基于不完备决策表的属性约简定义有多种,现研究基于知识粒度的属性约简。研究发现,差别矩阵是一种较好的设计属性约简算法的方法。为此,定义了一种粒度差别矩阵和基于该差别矩阵的属性约简,并证明了该差别矩阵的属性约简定义与基于知识粒度的属性约简定义等价。在此基础上,设计了一个新的基于信息量的不完备决策表的属性约简算法,其时间复杂度得以降低。There are many attribute reduction definitions.We researched attribute reduction based on knowledge granuation in incomplete decision table.The discernibility matrix method is a good way to design attribute reduction algorithm.So we gave the definition of discernibility matrix of granulation and the corresponding definition of attribute reduction.At the same time,we proved that the definition is the same as the definition about attribute reduction based on knowledge granulation in incomplete decision table.On this condition,we used the above discernibility matrix of granulation to design an efficient algorithm of attribute reduction based on knowledge granulation in incomplete.Its time complexity is reduced.

关 键 词：粗糙集 属性约简 不完备决策表 知识粒度 算法复杂度 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]