Abstract: Computing core attributes is one of the important research problems in rough sets theory. To overcome the shortcomings of computing core attributes based on discernibility matrix, the definition of decision discernibility matrix and method of dividing decision table are put forward. The methods of creating sub-decision discernibility matrix on sub-decision table and computing core attribute are presented, and it is proved that the core attribute is equivalent to the one based on positive region. The serial and parallel algorithms for computing core attribute based on horizontally partitioned decision table are designed. Both of the example analysis and experiment results show that the computing core algorithms proposed are correct and efficient.

Key words: rough set, positive region, decision discernibility matrix, core attributes

摘要： 核属性求解是粗糙集理论的主要研究内容之一。针对现有差别矩阵求核算法的不足，给出决策差别矩阵定义和水平划分决策表方法。提出在子决策表上创建子决策差别矩阵，进行核属性求解的方法;并证明了由该方法获得核与正区域核是等价的，同时设计相应的串行和并行求核算法。实例分析和实验比较表明所提出的求核算法是正确的、高效的。

关键词: 粗糙集, 正区域, 决策差别矩阵, 核属性