Modeling Cohesiveness and Enmity for Signed Social Communities

Social interactions can exhibit either positive or negative linkages, resulting in the abstraction of “signed graphs” or “signed social networks.” The social balance theory implies that a signed social network tends to converge to a balanced state where all individuals are divided into two cohesive groups that are antagonistic to each other. In this paper, we take this implication as a premise, and model the cohesiveness within each community and the enmity between two communities. We evaluate our model using synthetic signed networks, online signed social networks, and the U.S. Senate Voting network, and demonstrate its effectiveness in modeling the cohesiveness and enmity.