A typical social network is a single monolithic network to which individual nodes are initially attached. These nodes later become part of some existing groups or create new ones, eventually increasing the number of edges within the network. The task of detecting such groups (or communities, as they are commonly known) from such a social network is an interesting problem. The main challenge here is to determine the details of communities as accurately as possible. This research problem is often joined with another challenge: that of trying to develop better algorithms that optimize both time and space complexities for detecting such communities.
In this paper, the authors experiment by deploying a hybrid merging technique on small communities that are iteratively detected from the bottom up. These are later merged to reveal a larger well-knit community. The authors provide the outline for their algorithm, which they claim to be optimal insofar as time and space complexities are concerned. Furthermore, the complexity of the hybrid merging algorithm is compared with other popular algorithms to validate the claim.
A major criticism of studies on social networks could stem from the use of the term “community.” A community is often identified as a group that has certain intrinsic shared social practices. To equate groups present in a social network with communities could possibly be erroneous in the real sense of the term.