Automatic aspect discrimination in relational data clustering

Abstract

The features describing a data set may often be arranged in meaningful subsets, each of which corresponds to a different aspect of the data. An unsupervised algorithm (SCAD) that performs fuzzy clustering and aspects weighting simultaneously was recently proposed. However, there are several situations where the data set is represented by proximity matrices only (relational data), which renders several clustering approaches, including SCAD, inappropriate. To handle this kind of data, the relational clustering algorithm CARD, based on the SCAD algorithm, has been recently developed. However, CARD may fail and halt given certain conditions. To fix this problem, its steps are modified and then reordered to also reduce the number of parameters required. The improved CARD is assessed over hundreds of real and artificial data sets.