Abstract: In this paper, some discretization methods and their shortcomings are discussed and a new discretization method based on clustering is introduced; the method first clusters the samples according to their resemblance,then sets the discretized point in the attribute axle. Since the method takes the relevance of attributes into account, the more reasonable discretization result can be reached.

摘要： 分析了一些Rough Set中连续属性离散化的方法，指出了其中的某些不足，并给出了一个基于聚类的连续属性离散化的方法。对当前的论域中的例子根据相似性进行聚类， 对每个聚类在各属性轴上的投影的边界设离散断点。该方法考虑了各属性之间的相关性，能得到比较合理的离散结果。

关键词: Rough, set, 离散化, 聚类

CLC Number: 

• TP311.12