Abstract: In the covering generalized rough set theory, the original minimum description is primarily concerned with a single granulation. This paper first extends a single granulation to multi-granulation about the minimum description. A multi-granularity covering rough set model is established. Based on this minimum description, two different types of upper and lower approximation operators are proposed. Their corresponding  properties are studied and  a new algorithm for attribute reduction is given.

Key words: covering generalized rough sets, minimal description, multi-granulation, approximation operators, attribute reduction

摘要： 在覆盖广义粗糙集理论中，对最小描述的定义是建立在单一粒度基础上。将最小描述从单一粒度推广到多个粒度，建立了多粒度覆盖粗糙集模型。在此基础上，用最小描述建立了两类不同的上下近似算子，研究其性质，给出了一种基于最小描述下求属性约简的新算法。

关键词: 覆盖广义粗糙集, 最小描述, 多粒度, 近似算子, 属性约简