基于改进区分表的核属性约简算法  

A Core-Reduction Algorithm Based on an Improved Discernible Table

作　　者：程玉胜[1] 张佑生[1] 江效尧[2] 

机构地区：[1]合肥工业大学计算机与信息学院 [2]南京审计学院信息科学学院计算机系,江苏南京210029

出　　处：《计算机仿真》2007年第7期95-97,共3页Computer Simulation

基　　金：安徽省高校省级自然科学基金(2006kj040B)

摘　　要：知识约简是一个NP难问题,在众多知识约简方法中,核属性基本上作为属性约简的初始解,这样就能有效地减小约简算法在属性空间中的搜索范围,在一定程度上加速了知识约简的进行。然而,核属性解基本上是利用Hu的分辨矩阵法求得的。结合Hu的分辨矩阵算法,讨论了Hu以及Wroblewaski区分表算法中存在的问题,改进了区分表的定义并结合Ye方法提出了基于区分表核属性约简算法,既克服了分辨矩阵方法的缺陷,又避免了区分表算法中存在的问题。实验结果表明该算法能有效求出决策系统的属性核。Knowledge reduction based on rough sets theory is an important but NP - hard problem. Among the present various algorithms of knowledge reduction based on rough sets model, core is always initialized directly in knowledge reduction as attribute reduction set, which can efficiently reduce the searching scope of reduction algorithm in attributes space and accelerate the carrying out of knowledge reduction to some degree. However, the conclusion of core attribute is basically obtained by using Hu discernible matrix. The present paper, considering Hu＇s discernible matrix algorithm, is to discuss the problems existing in Hu and the algorithm of Wroblewaski discernible table, better the definition of discernible table and together with Ye method, and puts forward a core - reduction algorithm based on discernible table. In this way, it not only overcomes the disadvantage of methods of discernible matrix, but also avoids the problems in algorithm of discernible table. Result of experiment proves that this algorithm can effectively work out the attribute core in decision - making system.

关 键 词：粗集模型 知识约简 决策表 核 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]