Study On Urban Fire Based On Data Mining And GIS

With the accelerating process of urbanization,the elements of the city are becoming more complex and compact.The risk of urban disasters represented by fires has become more severe,and the number of occurrences has been increasing year after year.Unclear understanding of deeper unknown disaster-causing factors,unclear identification of catastrophe laws,and insufficient prevention and control methods are the main reasons why the existing fire protection system lags behind the rapid development of the city.Therefore,how to effectively excavate and ascertain the precursory information of the development of urban fires,to formulate reasonable fire control decisions,and advance the prevention and control of potential hazards is a difficult problem that needs to be solved urgently.Based on this,this paper uses the historical fire data of San Francisco as the research background,and the apriori algorithm is used to conduct in-depth analysis of the frequent item sets and association rules of urban fires,GIS technology is introduced to visualize the high-frequency fire item sets.Through constructing urban fire spatial correlation and fire occurrence prediction models,the main disaster-causing factors and evolutionary laws of urban fires are revealed from multiple time-space dimensions.The main results are as follows:(1)The Apriori algorithm is used to study the frequent item sets and association rules of urban fires,and the relationship between the causes and types of fires and places is analyzed through the method of correspondence analysis.The results show that urban fires mostly occur in fixed buildings.Considering the frequency and severity of fires,residential fixed buildings are dominant,and mechanical failure is the main cause of serious fires.The main causes of fire vary with different types and locations,among which improper use of materials and personnel error are the main causes of residential fire;electrical failure is the main cause of fire in places such as exhibition halls,and mechanical failure is the main cause of fire in places such as storage and outdoor tunnels.(2)Using GIS technology,the spatial distribution characteristics of urban fires are studied through the analysis methods of spatial point patterns such as nuclear density,clustering,and standard deviation ellipses.The results show that urban fires exhibit significant spatial agglomeration characteristics,and the central area of the agglomeration is in the northeast of the city.Fires of different attributes present different distribution characteristics in space,and the spatial agglomeration of fires caused by operator errors is the strongest;the spatial distribution of factory fires presents agglomeration characteristics and has obvious directionality.(3)A spatial correlation model of urban fires based on geographically weighted regression is constructed,and the spatial autocorrelation and heterogeneity of urban fires are studied.The results show that the spatial autocorrelation of urban fires is significant;the hot spots of fires in each year are in the northeast of the city,and there is no cold spot area.Factors such as fire station,POI,etc.contribute differently to the occurrence of fire in different areas,showing obvious spatial heterogeneity.(4)Through the method of data fitting,the law of fire alarm time is studied,the law of fire occurrence is analyzed from different time scales,the time series characteristics of the number of fires are studied,and a prediction model based on the ARIMA algorithm is constructed.The results show that the time of fire alarm is in accordance with the normal distribution,and there are certain differences in the time of fire alarm in different places.Among them,the time of fire alarm is the longest in storage and outdoor places.Urban fires show different characteristics on different time scales such as year,month,week,and time.January,weekends,and 1 pm-10 pm are the time periods with the most fires.A prediction model for the number of fire occurrences based on ARIMA was constructed,and it was found that the actual values are all within the upper and lower limits of the predicted values,and the relative error was between 16.67% and 32.18%.The research results of this paper expand the theory and methods of urban fire data analysis,and have important significance and application value for further revealing the occurrence and development of urban fires,constructing regional fire risk assessment models,guiding the allocation of urban fire protection resources and improving the urban fire prevention and control system.In this paper,46 figures,22 tables and 92 references are included.