A comparative study of the measures for evaluating community structure in bipartite networks

The community structure is an elementary property of complex networks. In bipartite networks, various global measures have been proposed to evaluate the quality of communities. However, few studies focus on comparing the performance of those measures on benchmark networks. Therefore, in this paper, we first propose a memetic algorithm (MACD-BNs) to detect communities in bipartite networks and then use it to separately optimize four existing measures to compare the performance of these measures. These four measures that belong to the same class are based on the idea that the same type nodes are grouped into a community due to their similar link patterns. Experimental results on some real-life and synthetic networks illustrate the effectiveness of the MACD-BNs. The comparison among different measures answers which of the four measures has the best relative performance in most cases and exposes the limitations of some measures.