A community discovery algorithm based on boundary nodes and label propagation

In this paper, a label boundary node algorithm (LBN) based on boundary nodes and label propagation is proposed. The algorithm reduces the update randomness in label propagation, which leads to the lack of stability in a community network. A core node can be found by traversing every node and calculating their values. Then, label propagation begins using the label to identify boundary nodes, so that the community division can be realized by giving weight to boundary nodes and deleting those boundary nodes by comparing their values. Finally, typical test data from the British college student clubs network dataset and the American university football game dataset are acquired. We also conducted an experimental comparison on the algorithm between the LBN and a label propagation algorithm (LPA). The comparison result demonstrates that the proposed algorithm is efficient, and that community division becomes more stable when the LBN algorithm maintains the original time efficiency.