Community detection in bipartite networks: Algorithms and case studies

Abstract

There is increasing motivation to study bipartite complex networks as a separate category and, in particular, to investigate their community structure. We outline recent work in the area and focus on two high-performing algorithms for unipartite networks, the modularity-based Louvain and the flow-based Infomap. We survey modifications of modularity-based algorithms to adapt them to the bipartite case. As Infomap cannot be applied to bipartite networks for theoretical reasons, our solution is to work with the primary projected network.We apply both algorithms to four projected networks of increasing size and complexity. Our results support the conclusion that the clusters found by Infomap are meaningful and better represent ground truth in the bipartite network than those found by Louvain.