Finding community structure in complex network based on latent variables

A number of recent studies have focused on detecting community structure in complex network. We develop an algorithm to detect communities by treating nodes as random variables and deriving samples of these variables from adjacency matrix which includes topology information of network. Using factor analysis theory, we represent communities as latent variables and extract the related matrix to uncover the relationship between network nodes and communities. The algorithm proposed in this paper uses the related matrix to improve the testing accuracy and we also notice that the algorithm overcomes the resolution limit possessed by other modularity-based methods in a kind of network topology structure. Experiments in real-world networks reveal that it detects significant and informative community divisions compared with other classic methods.