基于条件熵和变论域方法的属性约简算法  

AN ATTRIBUTE REDUCTION ALGORITHM BASED ON CONDITIONAL INFORMATION ENTROPY AND VARIABLE UNIVERSE OF DISCOURSE

作　　者：左金平[1] 郭玉栋[2] 秦振基[1] 

机构地区：[1]晋中学院计算机科学与技术学院,山西晋中030600 [2]晋中学院网络中心,山西晋中030600

出　　处：《计算机应用与软件》2010年第11期121-123,共3页Computer Applications and Software

基　　金：山西省高校科技研究开发项目(200613048);山西省教育科学"十一五"规划课题(GH-06203)

摘　　要：深入分析了论域变化与决策表条件信息熵的变化原理,在此基础上提出一种新的决策表属性约简算法。该算法通过不断将决策表中一定属性集下协调的部分从论域中去除,把不协调的部分作为新的论域做进一步处理,逐步减少参与属性约简的对象数,可以在一定程度上降低求取属性约简的计算量,实验数据表明该算法是高效的。In this paper,the principium of variations of universe of discourse and conditional information entropy of decision table is elaborately analysed.Based on this,a new algorithm of attribute reduction of decision table is proposed.In this algorithm,the consistency part in certain attribute set of the decision table is incessantly expelled from the universe of discourse,while further processing is taken on those inconsistency parts as the new universe of discourse for gradually abating the number of objects taking part into the attribute reduction.In this way the computation complexity of attribute reduction can be abated to a certain extent.Experimental data show that this algorithm is efficient.

关 键 词：决策表 属性约简 条件熵 变论域 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]