决策表属性约简的相对划分粒度表示

Presentation of Relative Partition Granularity of Attributes Reduction for Decision Table

粗糙集理论认为知识就是分类.本文对知识的分类能力给予了量化,提出利用划分粒度来定量地表示知识的分类能力.在划分粒度概念基础上,针对决策表定义了相对划分粒度并研究了它的性质,相对划分粒度可以定量表示决策表的条件属性子集相对于决策属性的分类能力的强弱;最后证明了对一致决策表的属性约简来说,相对划分粒度表示与Pawlak提出的代数表示是等价的.

Knowledge and classifications are related together by the theory of rough sets which claim is that knowledge is deepseated in the classificatory abilities of human beings. In this paper, we firstly give a quantitative representation of the ability of knowledge＇s classification, and provide a novel representation for knowledge, that is, it can be expressed by partition granularity. Secondly, the relative partition granularity is defined, and its qualities are discussed, where the relative partition granularity can be used to descript the classification ability of conditional attributes relate to decision attribute. Finally, the equivalence between the algebraic representation and the relative partition granularity representation is proved for a consistent decision table.