Syoji Kobashi, Takehiko Matsuura, Yutaka Hata, Seturo Imawaki*, and Makoto Ishikawa*
*Ishikawa Hospital
Proc. of Proc. of First Vietnam Japan Symposium on Medical Imaging /
Informatics and Applications
pp. 23-29, Nov. 2001.
This study introduces rough set, which can theoretically represent
the roughness of information, into the clustering algorithm.
The proposed clustering method evaluates degree of dissimilarly
between objects by using the concept of value reduct. The
indiscernibility relation, namely clustering result, is obtained by
evaluating the degree of dissimilarly with a roughness index.
The performance of the algorithm was evaluated by applying it to
public databases in UCI Machine Learning Repository, and biomedical
informatics data. Both of the clustering results are evaluated by
comparing with the conventional methods. The proposed method can
also give various clustering results according to the roughness index.