摘要： 基于差别矩阵思想的属性约简算法需要求出决策表的差别矩阵，然而差别矩阵的求取不但费时而且占用大量的存储空间。为此，提出一种基于差别矩阵中非空对象个数的改进属性约简算法。在利用差别矩阵思想的同时不生成差别矩阵，并给出属性重要度的定义及其快速计算公式，只需要 和 就能计算出属性重要度。实例分析证明，该算法能节省计算时间，求出最小属性约简。

关键词: 粗糙集, 简化决策表, 差别矩阵, 属性约简, 完备算法

Abstract: The method of attribute reduction algorithm based on discernibility matrix needs to compute the discernibility matrix of the decision table. However, computing the discernibility matrix not only spends much time but also needs a large of memory space. In order to get over the defect and use the idea of discernibility matrix methods, this paper presents a new improved attribute reduction algorithm based on the number of all not null objects. At the same time, the definition of the attributes importance is presented and a quickly computed formula is provided. The formula only needs to compute and to compute the attributes importance. Example analysis shows that the time is reduced, and it illustrates the minimum attribute reduction of the new algorithm.

Key words: rough set, simplified decision table, discernibility matrix, attribute reduction, complete algorithm

中图分类号: 

• TP311.12