Evaluation Of Direct Economic Losses From Earthquakes In Yunnan Province Based On GA-BP Neural Network

Yunnan Province is located at the eastern edge of the collision zone between the Eurasian plate and the Indian Ocean plate,with frequent seismic activity,special topographic structure and wide distribution of seismic zones,and has high seismic risk.Secondly,many places in Yunnan Province are relatively backward in development,with low seismic performance of buildings such as houses and high seismic vulnerability.Once a destructive earthquake occurs,it will cause huge economic losses,and has the characteristics of "small earthquake and medium earthquake catastrophe".After a destructive earthquake,in order to better guide the rescue work,it is first necessary to understand the economic loss of the disaster area,so the rapid assessment of the direct economic loss of the earthquake is particularly important,which is of great significance for rapid response after the earthquake to participate in the rescue.This paper analyzes earthquake risk management from three aspects: disaster analysis,earthquake risk sharing subject,and earthquake insurance pilot,and argues that it is necessary to quickly assess the direct economic losses of earthquakes in Yunnan Province.By analyzing the influencing factors of direct economic loss of earthquakes,fully considering the representativeness and rapid availability of data,the destructive earthquake data above magnitude 5 in Yunnan Province from 1990 to 2021 were collected,a total of 77 sample data were collected,and six influencing factors such as magnitude,polar seismic area intensity,focal depth of epicenter,seismic fortification intensity,population density and per capita GDP were preliminarily selected.Finally,the magnitude,polar zone intensity,focal depth,population density and per capita GDP were selected as the influencing factors of direct economic losses.At the same time,due to the obvious nonlinear relationship between each influencing factor and direct economic loss,each influencing factor has discreteness and randomness.Therefore,this paper introduces the BP neural network with better self-adaptive and self-learning effects to fit the nonlinear relationship between the direct economic loss of the earthquake and various influencing factors.However,because the BP neural network has the disadvantages of randomly unstable weights and thresholds,and it is easy to reach the optimal state at the local minimum point,so this paper introduces the genetic algorithm to optimize the BP neural network,enhance the ability of the global optimization of the neural network,and then achieve Optimize the effect of weights and thresholds.In order to avoid over-fitting,this paper divides the sample data into training set,confirmation set and test set,each accounting for 80%,10%,and10% respectively,and repeatedly adjusts the network parameters and genetic algorithm parameters,and then respectively adjusts the BP The neural network and GA-BP algorithm are used for cyclic training simulation,and the average relative error is selected to evaluate the prediction effect of the two.The empirical results prove that the average relative error between the actual output value of the GA-BP algorithm and the real economic loss value is smaller than that of the BP neural network.Considering that the intensity of the polar seismic area needs to be obtained after the intensity distribution map is published,this paper fits the relationship between the intensity of the polar seismic area and the magnitude and source depth,so as to quickly obtain the polar seismic zone intensity by fitting the relationship,and then uses the GABP algorithm to simulate and test the five seismic cases between the selected magnitude6-7,and finally finds that there is a good consistency between the real value of direct economic loss and the estimated value.This shows that this method can be used in the rapid assessment of direct economic losses of earthquakes to a certain extent,and has certain application value.