基于扩展可辨识矩阵的混合决策系统属性约简

经典粗糙集理论的研究对象只能是完备的离散决策系统。为了直接对不完备混合决策系统进行属性约简,因此通过引入邻域关系和限制容差关系的概念对可辨识矩阵的定义进行了扩展,提出了一种基于扩展可辨识矩阵的属性约简算法;该算法可以兼容处理完备与不完备混合决策系统。通过UCI数据集的仿真实验证明了该算法的有效性,最后讨论了扩展可辨识矩阵中的邻域阈值选择对属性约简结果的影响。

Attribute Reduction of Hybrid Decision System based on Expanded Discernibility Matrix

Classical Rough Set Theory,introduced by Professor Pawlak in 1982,its research object could only be the complete discrete decision system.In order to reduct the attribute of the incomplete hybrid decision system directly,expanded the discernibility matrix through the introduction of the concept of neighbourhood relation and limited tolerance relation,proposed a attribute reduction algorithm based on expanded discernibility matrix.The attribute reduction can either deal with incomplete or complete hybrid decision system.Demonstrated the effectiveness of the algorithm by simulating experiments of UCI data sets,concluded with a discussion on the influence of selection for neighbouring threshold of expanded discernibility matrix on attribute reduction results.