Community detection based on gravitational coefficient in collaboration network
Abstract
One important characteristic of complex networks is community structure. How to effectively divide the potential community structure of complex networks has been the focus of scholars because communities may have very different properties than the network. A community is usually defined as a collection of nodes with similar attributes. Generally, nodes in the same community are relatively densely connected to each other, compared with nodes from different communities. From the perspective of clustering, nodes in the same community can be considered as having higher similarities. Therefore, using graph clustering algorithms for community detection is theoretically feasible. Collaborative networks are special complex networks. A collaborative relationship tends to connect to multiple collaborators, which makes it hard to build collaborative networks by abstracting the collaboration into edges. Based on characteristics of the collaborative network, we expand the cluster similarity index and propose a gravitational coefficient index to measure the similarity of nodes and subsequently design community detection algorithms. Experiments using real datasets show that the proposed algorithm can obtain higher quality community partitioning results and avoid falling into local optimal solutions to obtain larger-scale communities than classical community detection algorithms.
