To study the relationships of the cores of attribute reduction definitions based on Hu's discernibility matrix (Hcore( C)), information entropy (Ecore (C) ), distribution ( Fcore ( C)) , maximum distribution (Mcore ( C) ), approximate (Acore ( C)) and positive region(Pcore( C)) , we proposed the different computing methods of the simplified discernibilitiy matrix for the above different cores. And we proved the following conclusions with the above methods that there are Hcore( C) ∪ Ecore( C) = Fcore( C) ∪ Acore (C) ) ∪ Pcore(C) and Hcore(C) ∪ Ecore( C) = Fcore (C) ∪ Mcore( C) in the inconsistent decision table. An example was used to illustrate that there are no relationships between Mcore(C)) and Acore(C)) as well as Mcore(C)) and Pcore(C)) in the inconsistent decision table. On the other hand, the above different cores are equivalent to each other in the consistent decision table.%为研究基于HU差别矩阵属性约简的核(HCore(C))、信息熵属性约简的核(ECore(C))、分布属性约简的核(FCore (C))、最大分布属性约简的核( MCore (C))、近似属性约简的核(ACore (C))和正区域属性约简的核(PCore (C))之间的关系,本文给出了上述各种核的简化差别矩阵计算方法,并用上述各种核的计算方法证明了如下结论:在不一致决策表中,HCore( C)(∩)ECore( C):FCore( C) (∩)ACore( C)) (∩)PCore( C);用实例说明了MCore(C))和ACore(C))之间没有关系,MCore (C))和PCore (C))之间没有关系;而在一致决策表中,各种不同核是相互等价的.
展开▼