Recent Approaches of Partitioning a Set into Overlapping Clusters, Distance Metrics and Evaluation Measures

Abstract

This paper reviews recently proposed overlapping co-clustering approaches and related evaluation measures. An overlap captures multiple views of the partitions in data set, hence is more expressive than traditional flat partitioning approaches. We present a graph-theoretic formulation of co-clustering which allows nodes to possess multiple memberships and hence finds usage in diverse applications like text mining, web mining, collaborative filtering and community detection. We also study proposed quality measures specifically adjusted to overlapping scenarios.particular subject.