The description of a set of objects can be done by various attribute sets. The selected attribute set implies relationships between object pairs. The resolution (comprehensiveness) of the relationships is affected by the selected attribute set. For a given resolution level (in terms of fuzzy application, resolution level is “level of fuzziness”) and a threshold value for the relationships, a partition can be defined for the objects. This problem is described and analyzed using graph theory notation. A reliability measure is suggested to judge various partitions. The results of the paper apply to fuzzy multivalent relations.
The presentation is nice. However, it assumes mathematical maturity on the part of the reader. The car-buyer example given in the introduction is a good motivation for the reader. However, it is hard to put the work into perspective since the author does not provide adequate references or a review of the previous work on the same subject.
Because of the purely theoretical approach, the paper will not be useful for people who are looking for a recipe for their clustering problem. It is valuable for researchers and graduate students. Another paper that is co-authored by Goetschel might also be helpful to the interested reader [1].