Abstract:

We proposed a local community detection algorithm based on topological potential, which uses topological potential of nodes to calculate their influence, and then takes the strategy of label propagation algorithm to detect overlap community structures via a new measurement index based on the similarity of local structures. The algorithm solves the problems of parameter setting, random result and high complexity of traditional algorithms. Algorithm comparison experiments on real world and computer generated datasets show that it is efficient.

Key words: community structure, topological potential, local similarity, label propagation, overlapping community