Aimed at the problems that reconnaissance-strike UAV is complex and diverse in combat tasks, difficult to obtain test data, and strong subjectivity in safety evaluation, an improved Bayesian network model is proposed to evaluate the deploying safety of the reconnaissance-strike UAV. According to the operational characteristics of the reconnaissance-strike UAV, five operational stages are divided, the safety risk factors in the use of different stages are identified, and the deploying safety evaluation index system of the reconnaissance-strike UAV in the strong confrontation environment is constructed. Based on the index system, the topological structure of Bayesian network is determined. The entropy method is utilized for improving the G1 method to obtain the prior probability of the root node, and the EM algorithm for obtaining the conditional probability of the child node. Finally, the data is imported into the GeNIe software for simulation, and the probability distribution of different use safety levels of econnaissance-strike UAV is obtained. The reverse reasoning of the model is carried out to identify the key factors that lead to the safe ty accidents of the reconnaissance-strike UAV.