A New Approach to Detect Important Members that Create the Communities in Bipartite Networks

The world around us consists of objects that have different relationships with each other. The result of these communications is various networks, part of which are bipartite networks. While many studies have investigated essential network members, less attention has been paid to the bipartite graphs. On the other hand, one of the most critical aspects of network analysis is the detection and extraction of communities that arise in the structure of networks. For these reasons, we have introduced a measure called H.H to identify influential nodes in community formation in the one-mode projection of a bipartite graph. The three main parameters that influence this measure are the size of the formed community, the effect of each node in the formation of that community, and the number of communities in which the node had an impact. The results of this paper show the differences of this measure with other centralities (eigenvector centrality, closeness centrality, betweenness centrality, and degree centrality) and how this measure takes into account aspects that other centralities do not. Through H.H score, we can find essential nodes that have been effective in forming a community, and by removing these nodes, communities can be eliminated. Any of the existing centralities has not addressed this issue, and this measure has sufficient independence to represent the important nodes in the formation of the communities. Experimental validation of the proposed measure is carried out on two real-world datasets: Southern Women Network and Person-Crime Network. The results of the implementation of the H.H score on the Person-Crime dataset show that by eliminating the nodes with the highest H.H score (top-10%), 29% of the communities have changed; this is while the centralities change the average of 18% of the communities and this shows the importance of the H.H score.