Taking the guest roller in Qiongzhou strait as the research object, the paper analyzes and summarizes the main influence factors to the safety state of guest roller. Take these factors as the nodes to build the Bayesian network topology of guest roller safety assessment model. Obtain the range of evidence nodes with Experts Mark and the nodes probabilities table by fuzzy inferences. then the fuzzy Bayesian network model of guest roller safety assessment is established. Finally, construct the guest roller safety index with Bayesian evidence reasoning. The safety index can represent security situation and trend of guest roller more accurately and dynamically.