Fault localization has theoretically been proven to be NP-hard. This paper takes a weighted bipartite graph, as fault propagation model, and proposes a heuristic fault localization algorithm based on Bayesian suspected degree (BSD) to reduce the computational complexity. It introduces a metric of BSD, which needs only to be calculated once, and uses incremental coverage, which makes the algorithm a low computation complexity O(|F|×|S|). Simulation results show that the algorithm has a high fault detection rate as well as low false positive rate and performs well even in the presence of unobserved and suspicious alarms. The algorithm, which has a polynomial computational complexity, can be applied to a large-scale communication network.