Multiscale community estimation based on temporary local balancing strategy

Abstract

Community division in complex networks has become one of the hot topics in the field of network science. Most of the methods developed based on network topology ignore the dynamic characteristics underlying the structure. By exploring the diffusion process in the network based on random walk, this paper sums up the general rule with temporal characteristics as a temporary local balancing strategy which can be used in the community division. The strategy divides the network into different communities according to the duration of a stable local balancing state in the diffusion process. The longer the duration, the more stable the structure of the community in that state. Applying the strategy to computer-generated and real-world networks, respectively, it is proved that these temporary local balancing states existing in the diffusion process can reveal the internal community structure of the network. In addition, the modular structure appears at different time scales of diffusion process, similar to the hierarchical organization, and also provides a new perspective for multiscale network community detection.