A Rough Set Based Clustering Method by Knowledge Combination

Tomohiro OKUZAKI  Shoji HIRANO  Syoji KOBASHI  Yutaka HATA  Yutaka TAKAHASHI  

IEICE TRANSACTIONS on Information and Systems   Vol.E85-D   No.12   pp.1898-1908
Publication Date: 2002/12/01
Online ISSN: 
Print ISSN: 0916-8532
Type of Manuscript: PAPER
Category: Databases
rough set,  clustering,  knowledge combination,  

Full Text: PDF(376.9KB)>>
Buy this Article

This paper presents a rough sets-based method for clustering nominal and numerical data. This clustering result is independent of a sequence of handling object because this method lies its basis on a concept of classification of objects. This method defines knowledge as sets that contain similar or dissimilar objects to every object. A number of knowledge are defined for a data set. Combining similar knowledge yields a new set of knowledge as a clustering result. Cluster validity selects the best result from various sets of combined knowledge. In experiments, this method was applied to nominal databases and numerical databases. The results showed that this method could produce good clustering results for both types of data. Moreover, ambiguity of a boundary of clusters is defined using roughness of the clustering result.