Clustering using similarity based on uniqueness measure and its properties
Manami Matsumoto, M. Emoto, Masao Mukaidono · 2005
Various similarity relations have been proposed until now dealing with number of coincided attributes, Euclid distance, etc. In 2003, we have proposed similarity based on uniqueness measure. The similarity is based on human's perception and is obtained by using uniqueness of attribute values in a subset of all objects in an information system. We regard the subset as knowledge. In the similarity, it is possible to change order of similarities by knowledge. For example, in certain knowledge, object A is more similar to object B than object C. In the other knowledge, object A is more similar to object C than object B. We consider clustering using similarity based on uniqueness measure and its properties.