A Network Embedding-Enhanced Approach for Generalized Community Detection

Community detection is one of the most important tasks in network analysis. Many community detection methods have been proposed recently. However, they typically focus on assortative community structures (i.e. nodes within the same community have more connections), while ignoring the diversity of community patterns in real world. In addition, the network topology, which these methods are mainly based on, is often noisy and very sparse. These two issues bring difficulties to existing methods for accurately finding communities. To address these problems, we propose a new probabilistic generative model. In this model, we first use an idea of mixture modeling to describe network regularities, and then introduce network embeddings to further enhance the ability of this model to describe network communities. Based on these, the new model will not only find generalized communities (e.g. assortative communities, disassortative communities, and their mixture), but also be robust for community detection in complicated situations (e.g. on very sparse networks with large noise). We present an efficient expectation-maximization (EM) algorithm to learn the model. Finally, we demonstrate the superior performance of our new approach over some state-of-the-art methods on both synthetic and real networks, and also validate its robustness to the above issues via a case study analysis.