In this dissertation, we propose novel Bayesian machine learning models to solve various graph analytics problems, including graph representation learning, graph generative modeling, structured semi-supervised learning, ...