Abstract: With the fast development of information and communication technology, our society has entered the era of big data. How to dynamically update attribute reduction is vital to the efficiency of knowledge discovery. A general heuristic attribute reduction algorithm consumes a great deal of computational time. These reduction algorithms are inefficient to deal with the large-scale data. This papers proposes a matrix-based attribute reduction approach under multi-granulation rough set. Experiments have been performed two data sets from UCI, and the results validate that the proposed attribute reduction approach under multi-granulation rough set can achieve better performance for large-scale data sets.

Key words: rough set, multi-granulation, attribute reduction, knowledge granularity

摘要： 随着网络和通信技术的快速的发展，社会进入了大数据时代。如何能够快速地从海量大数据中找到属性约简是目前研究的一个热点。由于传统属性约简的方法在计算大数据属性约简时，需要消耗巨大的计算时间，不能有效地处理日益积累的大数据属性约简的问题。为了提高传统属性约简算法的效率，针对较大决策信息系统属性约简更新问题，利用多粒度粗糙集理论，提出了基于多粒度粗糙集模型的矩阵属性约简算法，通过2组UCI数据集对所提出的多粒度矩阵属性约简算法的性能进行测试，结果验证了该多粒度矩阵属性约简算法是合理且有效的。

关键词: 粗糙集, 多粒度, 属性约简, 知识粒度