Granular meta-clustering based on hierarchical, network, and temporal connections

In granular computing, each object is represented as an information granule and an information granule can be connected to other granules through semantic relationships. These connections can lead to a granular hierarchy or a network. Data mining of one set of objects may not be able to capture information contained in granular connections. This paper describes a concept of meta-clustering that clusters a set of granules using clustering information from another or the same set of networked granules. Cluster membership of one granule can affect another granule's cluster membership, resulting in a recursive meta-clustering process. We illustrate the usefulness of such meta-clustering for a granular hierarchy consisting of sets of businesses and reviewers, a set of networked granules representing mobile phone users, and trading patterns of financial instruments that are linked to each other through a temporal dimension.