差别矩阵浓缩及其属性约简求解方法

属性约简是粗糙集理论的重要研究内容之一，已出现大量的属性约简算法，其中基于差别矩阵的属性约简算法是高效属性约简算法之一，但这些算法主要针对一致决策表，而对于不一致决策表，某些情况下不能得到属性约简。为此，本文提出改进的差别矩阵及其属性约简求解方法，统一考虑决策表一致和不一致情况两种情况下的属性约简，有效改进经典的基于差别矩阵求解属性约简的不足。同时，为适应大数据集属性约简需要，提出一种新的差别矩阵浓缩策略，以此提高属性约简的效率。

Discernibility Matrix Enriching and Computation for Attributes Reduction

Attributes reduction is one of important parts researched in rough set theory. Thus, many algorithms have been proposed for attributes reduction, in which the algorithms based on discernibility matrix is one of efficiently attributes reduction algorithms. Unfortunately, these algorithms based on discernibility matrix mainly aim at the consistent decision table, and can not get a correct result for an inconsistent decision table in some cases. Therefore, in this paper, we introduce improved discernibility matrix for computing attributes reduction, which gives an unified framework for a consistent or inconsistent decision table, and efficiently improves the drawback of the existing attributes reduction algorithm based on discerniblity matrix. At the same time, a novel method of improved discernibility matrix enriching is proposed for attributes reduction of a very large dataset.