ASAClu: Selecting Diverse and Relevant Clusters

Abstract

No clustering algorithm is guaranteed to find actualgroups in any dataset. To deal with this problem, one can applyvarious clustering algorithms, generating a set of partitions andexplore them to find the most appropriated ones. The number ofpartitions and its component clusters may be too large, making itdifficult to the specialist analyze the final result. In this paper, weintroduce a new selection strategy namedASAClu, which is aimedat selecting a reduced set of clusters from a given collection ofpartitions generated by different clustering algorithms.