Improved Modularity Based on Girvan-Newman Modularity

Social networks can be modeled by graphs with nodes and edges, and communities are sub graphs within networks - groups of nodes within which connections are dense, but between them connections are sparser. According to this property of communities, this paper proposes a new modularity for measuring how good a particular division is based on the concept of coupling coefficient. Further more, this paper applies proposed modularity to synthetic network data and compares the computational results under different modularity. The experimental results show that our new modularity is suitable for the cases that all communities have nearly the same number of links, and it is also suitable for the cases that the number of links in a community differs greatly from the one in another community.