Abstract: In order to strengthen the early warning and management of fire risks at hotel sites, the China Fire Yearbook and the fire data of Dongguan City in Guangdong Province were analyzed and summarized, and relevant literatures were consulted. A hotel fire risk model based on Bayesian networks was established and weighted the ratio method was used to calculate the probability of a hotel  fire in Dongguan. The software Genie 2.1 was used to model the fire risk factors. Sensitivity analysis of each fire risk was performed to identify the factors that have a significant impact on the fire risk. Suggestions for improving the hotel's fire risk were put forward for the significant factors.

Key words: fire safety, weight analysis, Bayesian Network, hotel fire