A dynamic evolutionary clustering perspective: Community detection in signed networks by reconstructing neighbor sets

• A community detection algorithm is presented for signed networks.
• The detection performance is higher than two recent community algorithms for signed networks and the runtime is reduced.
• Differential equations are proposed to imitate the constantly changing states of the nodes in signed networks.
• The main process of the evolutionary clustering algorithm is based on the reconstructed neighbor sets.
• The analytical results are verified by comparative experiments on both synthetic and real world networks.

Community detection in social networks has been intensively studied in recent years. In this paper, a novel similarity measurement is defined according to social balance theory for signed networks. Inter-community positive links are found and deleted due to their low similarity. The positive neighbor sets are reconstructed by this method. Then, differential equations are proposed to imitate the constantly changing states of nodes. Each node will update its state based on the difference between its state and average state of its positive neighbors. Nodes in the same community will evolve together with time and nodes in the different communities will evolve far away. Communities are detected ultimately when states of nodes are stable. Experiments on real world and synthetic networks are implemented to verify detection performance. The thorough comparisons demonstrate the presented method is more efficient than two acknowledged better algorithms.