摘要
根据决策表定义条件属性区分能力指数DI(a)的概念,给出DI(a)的若干性质。定义拟等价类的概念,提出基于区分能力指数的决策表相对约简算法,该算法基于分治策略,用属性a代替相应区域的区分元素,从而减小算法搜索的空间。实验结果表明,该算法具有较高的搜索效率,能够得到全部或大部分约简。
Discernibility ability index DI(a) of decision table is defined,some properties about DI(a) are presented.Concept of quasi-equivalence class is defined,a relative reduct algorithm about decision table is proposed based on the discernibility ability index.The algorithm uses divide and conquer strategy to reduce research space by replacing discernibility element with attribute a.Experimental results show that the algorithm has better efficiency and can obtain all or majority of reducts.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第17期38-39,43,共3页
Computer Engineering
基金
国家科技支撑计划基金资助项目(2006BAK01A33)
吉林省科技发展计划基金资助项目(20070321
20090704)
关键词
数据挖掘
粗糙集
约简
区分矩阵
区分能力指数
data mining
rough set
reduct
discernibility matrix
discernibility ability index