不完备信息系统中基于限制容差关系的属性约简方法

决策表核属性的确定往往是信息约简的基础，然而以往的核属性约简方法大多是针对完备信息系统的。将完备信息系统中的属性核与属性序约简算法延伸至不完备系统，提出一种不完备信息系统中基于限制容差关系的属性约简方法。该方法通过构造限制容差关系下决策表的改进分辨矩阵来求得核属性，并将非核属性按直观影响分类质量的能力排序，能够保证得到的约简结果是相对最小约简。通过实验比较证明该方法可行、有效。

Attributes reduction method based on limited tolerance relation in incomplete information system

The confirm of core attribute of a decision table is always the base of information reduction. However, most of the previous reduction methods in core attribute were faced to complete information system. Extending the reduction algorithm in core attribute and attribute order to incomplete information system, which used in complete information system, this paper presents a attribute reduction algorithm based on limited tolerance relation in incomplete information system. This method obtains core attribute by constructing a improved discernable matrix of decision table in incomplete information system, sorts the attribute which not belongs to core attribute by ability of affecting classification quality intuitively, in order to ensure the reduction result we get is a relatively minimize reduction. This method shows more feasible and effective in experiment by comparison.