Abstract: The data on construction safety risk accidents in China in recent years are collected and sorted out. The main risk factors in Xi'anlu station of the Dalian subway phase Ⅰ project are determined by the expert investigation method. According to the causal relationship among risk factors, the corresponding Bayesian network model is determined. The construction risk occurrence probability of this project can be predicted by using forward causal inference technology based on the Bayesian network, and construction risk can be diagnosed by using reverse inference technology. The results show that the risk probability of the construction safety of this subway project is 0.530. The data shows that the overall risk level of construction safety of this subway project is medium. The key factors that make the Xi'anlu station of the Dalian subway phase Ⅰ project in a dangerous state are inaccurate construction measurement data, non-detailed construction organization or construction scheme, and unreasonable connection of the operation process.