Abstract: Based on the monthly fire alarm data of super large cities, combined with the relevant data of climate, economy and so on, this paper constructs the influencing factor model of the monthly distribution of the number of fires, and analyzes the explanatory power of climate and economic variables on the occurrence of fires through the negative binomial regression of counting data. The results show that on the premise of controlling calendar factors, the combination of independent variables of rainfall days and total industrial output value has a good explanatory power for the monthly fire alarm times, and the data of temperature, electricity consumption and so on are less important for the fire alarm number. Under the condition that other variables remain unchanged, the fire alarm decreases by 2.1% for each additional day of rainfall; For every 100 million yuan increase in the total industrial output value, the fire alarm will increase by 0.033%; For every increase in the number of holiday, the fire alarm will increase by 2.4%.

Key words: number of fire alarms, monthly distribution, economic factors, climatic factors, negative binomial regression, count data