Abstract:
Community detection plays an essential role in understanding network topology and mining underlying information. A bipartite network is a complex network with more important authenticity and applicability than a one-mode network in the real world. There are many communities in the network that present natural overlapping structures in the real world.
The starting point of this paper is to study and analyze two different community detection algorithms, BCFinder and Panda+. From this analysis some major problems are being pointed out, and then I propose a new algorithm, merging the main ideas of the methods analyzed, to efficiently mine overlapping communities in large-scale sparse bipartite networks. It only depends on the network topology and does not require any prior knowledge about the number or the original partition of the network.
Finally, a wide analysis is proposed divided in two main parts: first I compare my algorithm with four other state-of-the-art methods, based on a bipartite approach, to have a deep view over a range of different solutions and to show that my algorithm has better results with some specific conditions, using artificial and real-world networks with ground-truth information, to evaluate the quality between the chosen methods.
In the last part, considering a bipartite network, I analyze the differences between the bipartite and unipartite approach, one that works directly on the network and the other that need to pass by a one-node transformation. Experimental results show that it is better to consider working directly with the bipartite methods, to not loose many information due to the transformation used by the unipartite approach.
