A community integration strategy based on an improved modularity density increment for large-scale networks

This paper presents a community integration strategy for large-scale networks, based on pre-partitioning, followed by optimization of an improved modularity density increment ΔD. Our proposed method initially searches for local core nodes in the network, i.e. potential community centers, and expands these communities to include neighbor nodes which have sufficiently high similarity with the core node. In this way, we can effectively exploit the information of the node and structure of the network, to accurately pre-partition the network into communities. Next, we arrange these pre-partitioned communities according to their external connections in descending order. In this way, we can ensure that communities with greater influence are prioritized during the process of community integration. At the same time, this paper proposes an improved modularity density increment ΔD, and shows how to use this as an objective function during the community integration optimization process. During the process of community consolidation, those neighbor communities with few external connections are prioritized for merging, thereby avoiding the fusion errors. Finally, we incorporate global reasoning into the process of local integration. We calculate and compare the improved modularity density increment of each pair of communities, to determine whether or not they should be integrated, effectively improve the accuracy of community integration. Experimental results show that our proposed algorithm can obtain superior community classification results on 5 large-scale networks, as compared with 8 other well known algorithms from the literature.