Reserch On Fire Risk Assessment Model Of City Area Based On The BP Neural Network

With the acceleration of urbanization and the increasing severity of urban fires,the study of urban fire risk assessment has become an important issue.In order to make a more objective assessment of the safety risks of fires in urban areas and to predict regional fire risks,this paper uses 22 districts and counties in Chengdu as the target area to establish an urban area fire safety risk assessment index system.At the same time,collect all the data corresponding to the indicator and quantify the data and assign the risk.Firstly,the evaluation results are obtained by the analytic hierarchy process and a verification model of regional fire safety risk assessment based on BP neural network is established to verify the feasibility of BP neural network for regional fire risk assessment.Then,the regional fire risk is represented by the economic loss rate caused by the fire in the urban area,and the fire economic loss rate is graded and assigned according to certain standards.Based on this,an urban fire risk assessment and prediction model based on BP neural network is established.The index data of the index system is used as the input of the network model,and the value of the fire economic loss rate is taken as the output of the network model.Simultaneously,Fire data from various districts and counties in Chengdu were used as model training samples and test samples to acclimate and test the models,and models were continuously adjusted to optimize the model to obtain urban fire safety risk assessment and prediction models based on BP neural network.Finally,in order to study the impact of the number of indicators on the accuracy of the model,according to the “contribution” of the assessment index to the fire risk,80%,85%,90%,and 95% of the four indicators are used as sample input to train the model and compare the accuracy of the model.In the end,the following four main conclusions are drawn:(1)By using all indexes of the evaluation index system as sample input and the evaluation value obtained by the analytic hierarchy process as the sample output to acclimate the neural network model,the average prediction accuracy of the final network model is 97.17%.The neural network training belongs to the “tutor” algorithm,and the reliability of the output data that plays a guiding role is high.The model established with the neural network can quickly and accurately obtain reliable assessment results.The 97.17% model accuracy verifies the feasibility of using BP neural network model to assess regional fire safety risks.(2)By studying the influence of different hidden layer nodes on the prediction accuracy of the neural network model,we can find that as the number of network nodes increases,the accuracy of the assessment model increases and then decreases.This proves that the excessive number of nodes may lead to the model.This verifies that too few nodes may lead to insufficient learning of the model,and excessive number of nodes may lead to over-learning of the model.Only in the right hidden layer network,the prediction accuracy of the nodes will be the highest,and the performance of the model will be optimal.(3)Based on the evaluation index system index as a sample input and the economic loss rate evaluation score as a sample output to acclimate the neural network model,the average prediction accuracy of the final network model is 95.03%.This shows that when the regional economic fire index is selected to represent the fire risk in the region,the method of using neural networks and the regional fire safety risk assessment index system established in this paper can predict the fire in the target area with high precision.The loss rate situation can be compared horizontally across the country,which helps the region to macro-control its own level of fire risk.(4)When adopting 80% evaluation index as a sample input,the accuracy of the network model has reached more than 90%.With the increase in the number of input indicators,the prediction accuracy of the domesticated network model is continuously improving,and all the quantitative indicators are used as the sample input,the network error minimum model prediction accuracy is the highest.This has a certain guiding role in the determination of assessment indicators during fire risk assessment.Any index related to fire risk can be used as a sample input for neural networks.Even in the process of increasing the number of indicators,there may be some indicators are not independent of each other,but the neural network model has a higher robustness and fault-tolerance performance,This proves that using the BP neural network model to assess the superiority of regional fire safety risks,this superiority is that it can guarantee the results of the assessment of the iterative and increasing indicators,which is of great significance to the regional fire safety risk assessment.Repeated information and even error messages for input indicators have a high degree of inclusiveness and will not cause much impact on the results.This proves that using the BP neural network model to assess the superiority of regional fire safety risks,this superiority is that it can guarantee the results of the assessment of the iterative and increasing indicators,which is of great significance to the regional fire safety risk assessment.