基于相对知识粒度的决策表约简

Reduction for decision table based on relative knowledge granularity

知识粒度主要应用于信息系统的属性约简。为了把知识粒度拓展到决策表约简领域,在研究知识粒度的基础上,定义了相对知识粒度的概念,证明了对一致决策表约简而言,相对知识粒度表示与Pawlak代数表示的等价性。进一步定义了基于相对知识粒度的属性重要度,提出了两个基于相对知识粒度的启发式决策表约简算法。通过理论分析与实例表明约简算法是有效可行的。

The knowledge granularity was mainly used for attribute reduction in information systems. In order to expand the knowledge granularity to the field of decision table, the relative knowledge granularity was defined based on knowl- edge granularity. The equivalence between the Pawlak algebraic representation and relative granularity representation was proved for a consistent decision table. Based on the definition of relative knowledge granularity, the attribute signif- icance was defined, and two heuristic reduction algorithms for decision table were proposed. Theoretical analysis and the actual example study showed that the reduction algorithms were efficient and feasible.