Hierarchical Clustering of Non-Euclidean Relational Data Using Indiscernibility-Level

In this paper, we present a clustering method for non-Euclidean relational data based on the combination of indiscernibility level and linkage algorithm. Indiscernibility level quantifys the level of global agreement for classifying two objects into the same category as indiscernible objects. Single-linkage grouping is then used to merge objects according to the indiscernibility level from bottom to top and construct the dendrogram. This scheme enables users to examine the hierarchy of data granularity and obtain the set of indiscernible objects that meets the given level of granularity. Additionally, since indiscernibility level is derived based on the binary classifications assigned independently to each object, it can be applied to non-Euclidean, asymmetric relational data.