Research On Fire Probability Of Urban Residential And Commercial Places Based On Bayesian Network

As an artificial environment with a large carrier,cities have the characteristics of complex functions,intensive personnel,property of concentrated.The rapid development of cities and the increasing number of disasters include frequent fire accidents and huge losses.The guide of urban disaster prevention systems Urban planning is less concerned about this aspect.Although the overall fires are frequent,a specific building fire is a random event with a small probability.At present,there are different methods to predict the probability of fire occurrence.The probability of collapse in a fire predicts fire damage.This paper predicts the number of fires and direct economic losses of urban fires,calculates the probability of fires in two types of buildings that are typical of urban fires,namely residential and commercial fires.The fire probability of commercial premises is obtained under Bayesian network reasoning,which mainly includes the following research contents:First,calculate the fire data of Foshan from 2009 to 2018,and build a prediction model based on the number and economic loss.The residual and relative residuals are obtained.The average relative residual is 0.087,the accuracy of the model is 91.33%,and the accuracy level of the model is At the first level,the number of fires in 2019 and 2020 in the next two years is 1301 and 1412,respectively.In the same way,a model of fire loss is established with a model accuracy of 96.775% and a grade of one.The direct fire loss predictions for 2019 and 2020 are 50.877 million yuan and 46.868 million yuan.Secondly,on the basis of extracting the statistical yearbook data,the characteristics of domestic and commercial fire accidents in domestic and commercial premises were analyzed,and the differences in fire numbers,direct economic losses,and casualties between domestic and foreign developed countries and commercial premises fires were compared.The proportion of fire disaster hazards in residential and commercial cities outside the city.Among the proportions,the first and second largest proportion of residential disaster hazards are electrical(37%)and accidental use of fire(28%).Disaster factors accounted for the first and second place are electrical(49%)and other fire causes(17%).It points out the fire probability of residential and commercial sites in various countries from 2007 to 2016.Domestic residential fires average 1.69 ×10-4 cases / household each year.,the maximum is 2.8 × 10-4 cases / household;the average annual fire probability of domestic commercial places is 5.19 × 10-5 cases / m2,and the maximum is 6.8 × 10-5 cases / m2.The probability of fire in a commercial place is significantly higher than the probability of a fire in a house.Due to different statistical methods,the number of commercial places is significantly less than that of a house.Third,analyze the fire and commercial site disaster factors,and use Bayesian analysis software Genie2.1 to build a Bayesian network model of fire and commercial sites in Foshan.Combine existing knowledge and machine learning to modify the network model.The data nodes assign initial probability values and perform a sensitivity analysis on the model to obtain a combination of fire factors that have a greater impact on residential fires.Among them,electrical,fire accidental,and smoking occurrences have the highest probability of residential fires,which is consistent with reality.At the same time,the probabilistic risk values of fires in residential and commercial places are 0.92 × 10-4 cases/ouseholds and 1.3 × 10-5 cases / m2,and the prevention and control measures for residential and commercial places are given.Finally,combined with the fire accident samples of residential buildings in Foshan City,the validity of the model is analyzed to verify the reliability of the model.The analysis shows that after adding fire evidence data and node status,the probability of residential fires is significantly increased,most of which are at the range of 40%-50%,it is significantly higher than the average residential fire in this city.