Host: Japan Society for Fuzzy Theory and Intelligent Informatics (SOFT)
In this paper, we present a method for representing the granularity of data that have non-metric relational properties. Asymmetric, relational proximity is transcribed into a set of binary classifications, and the indiscernibility of objects are characterized by Jaccard coefficients. Objects are merged into single granule when we disregard fine discrimination knowledge by changing indiscernibility level. According to this, we built a dendrogram that represented hierarchy of granules. Using the brand swith data, we demonstrate that the hierarchy of granules produced by the discernibility-based approach could have reasonable correlation with the original dissimilarity matrix.