To perform scientific and rational evaluation on landslide dam stability, essential information was undertaken from 313 collapsed and intact landslide dams worldwide to establish a case database. Eight influential factors related to the morphological characteristics and material composition of landslide dams, hydrodynamic conditions of barrier lakes were considered. Based on the analysis of Bayesian networks, two novel and expedited methods for assessing landslide dam stability were introduced, including a simplified model with missing nodes and a detailed model with intact nodes. The comparison between typical models of rapid evaluation method of landslide dam stability from both home and aboard shows that the simplified Bayesian network model can achieve an absolute accuracy rate of 83.33% and a conservative accuracy rate of 94.44%. Similarly, the detailed model demonstrates absolute and conservative accuracy rates of 80.77% and 96.15%, respectively. The reliability and scientificity of the Bayesian network model are verified. The Bayesian network models can incorporate and update real-time sample sets, and thereby the predictive accuracy of the models can be increased perpetually.