Community Detection using Fortunato's Performance Measure (2509.00938v1)

Abstract: In his paper on Community Detection [1], Fortunato introduced a quality function called performance to assess the goodness of a graph partition. This measure counts the number of correctly ``interpreted" pairs of vertices, i. e. two vertices belonging to the same community and connected by an edge, or two vertices belonging to different communities and not connected by an edge. In this paper, we explore Fortunato's performance measure (fp measure) for detecting communities in unweighted, undirected networks. First, we give a greedy algorithm fpGreed that tries to optimise the fp measure by working iteratively at two-levels, vertex-level and community-level. At the vertex level, a vertex joins a community only if the fp value improves. Once this is done, an initial set of communities are obtained. At the next stage, two communities merge only if the fp measure improves. Once there are no further improvements to be made, the algorithm switches back to the vertex level and so on. fpGreed terminates when there are no changes to any community. We then present a faster heuristic algorithm fastFp more suitable for running on large datasets. We present the quality of the communities and the time it takes to compute them on several well-known datasets. For some of the large datasets, such as youtube and livejournal, we find that Algorithm fastFP performs really well, both in terms of the time and the quality of the solution obtained.