Abstract:
The paper addresses the community membership problem of predicting whether two nodes belong to the same community or not,
using a hill climbing based breadth first traversal algorithm and a
probabilistic depth first traversal algorithm. The algorithms have
lesser time complexity as compared to the standard community detection algorithms. The proposed algorithms achieve a high efficiency in terms of the fraction of node pairs correctly identified,
which is greater than 70% in most of the network instances. Moreover, a very little fraction of the graph is traversed which is local to
the nodes in consideration.
