Abstract: In order to effectively prevent the occurrence of large commercial complex fire accidents, and avoid and reduce the damage and loss caused by fire to people to the maximum extent, this paper combines the Bow-tie (BT) model and Bayesian Network (BN) model to make a dynamic risk analysis of the fire accident probability of large commercial complexes. On the basis of historical data in recent years, the causes of fire accidents in large commercial complexes are identified by BT model, and the prior probability of basic events is determined by statistical data. Then, based on GeNie software, BT model is transformed into BN model according to logical transformation rules, and the established model is optimized according to the actual fire development stage. Taking a large commercial complex in Kunming as an example, the variables of each node are analyzed through the posterior probability and sensitivity analysis of nodes, and the key factors that induce the fire of the large commercial complex are extracted to verify the feasibility of the established BT-BN model, which is also convenient for the subsequent formulation of corresponding management and technical measures for the fire risks of the large commercial complex.

Key words: risk analysis, large commercial complex, fire, bow tie model, Bayesian network