基于覆盖粗糙集模型的层次聚类算法

目前大部分聚类算法只适用于处理属性取值为单值的数值型数据,介绍了一种新的基于粗糙集理论的聚类算法,该算法不仅可用于取值为单值的数值型数据聚类,而且能够用于取值为多值的非数值型数据聚类.该算法利用基于相容关系的属性最小覆盖来求解对象各属性的对象属性信息粒.在此基础上,通过对象属性信息粒和对象粗糙相似度的运算构建各对象的相容粒.最后,把具有相同相容粒的对象视为同一等价类,从而实现对论域的聚类,进而对数据对象进行层次聚类.实验结果表明,该算法是可行的.

Hierarchical clustering algorithm based on cover rough set model

Now most of clustering algorithm can only deal with single numerical value of data, a new clustering approach based on rough set is introduced, this clustering approach can be used not only for the single numerical value, but also can be used for the multiple non-numerical value. Firstly, based on attribute minimum covers and tolerance relation, object attribute information granulations of attributes of objects are generated. Secondly, tolerance granulations of objects are generated depending on the object attribute granulations and rough similarity between objects. Finally, the objects with same tolerance granulations consist of an equivalence class of the information system, and hierarchical clustering of the information system is implemented. The experiment shows that the proposed approach is feasible.