Research On Earthquake Emergency Material Demand Forecast Based On FOA-BP Neural Network

According to the unpredictable characteristics of natural disasters,natural disasters have caused great losses of personnel and property to human society.A rapid response to the supply of emergency materials is the key to reduce the loss of people’s lives and property.To realize the supply of emergency materials,it is urgent to make a scientific,reasonable and effective prediction of the demand for emergency materials.At the same time,efficient and timely emergency material demand forecasting can provide effective policy adjustment suggestions for decision-makers.Due to the frequent occurrence of earthquakes in China due to geographical factors,earthquake disasters,as one of the most serious natural disasters,usually cause inestimable losses to the society.The supply of materials to the disaster areas after the earthquake is the most important content of the rescue work.As the detailed information of the disaster areas after the earthquake can not be fed back in time,emergency departments can not judge the needs of specific materials,the amount of material distribution will lack accuracy.Excessive supply leads to serious waste,but the supply is too small to meet people’s needs for materials.Therefore,before the emergency department supplies to the disaster area,it is very important to predict the demand for emergency materials in the disaster area timely and correctly,which is not only conducive to disaster relief management,but also can provide reference for emergency departments to make decisions.In this dissertation,the indirect prediction method is used to predict the demand of emergency materials after the earthquake.firstly,for the prediction of the number of casualties,the BP neural network with three-layer structure is selected as the prediction model.BP neural network has a strong nonlinear mapping ability,so it can model the nonlinear problems of this dissertation relatively effectively,so as to better fit the number of casualties,and then forecast the emergency materials through the number of casualties.This dissertation points out that the traditional BP neural network has some shortcomings such as slow convergence and easy to fall into local optimization,and proposes to use principal component analysis to process emergency data in order to remove redundant information and reduce the input of BP neural network.Secondly,forest optimization algorithm is used to improve BP neural network,using the global search ability of forest optimization algorithm and the characteristics of avoiding falling into local optimization.The problems existing in BP neural network are further solved.It is proved by experiments that compared with the three prediction models of traditional BP neural network,PSO-BP neural network and FOA-BP neural network,the root mean square error,average absolute error and average absolute percentage error are used to evaluate the performance of the model,which verifies the effectiveness and accuracy of the FOA-BP neural network model.Finally,when estimating emergency supplies,the number of casualties and safety inventory theory are combined to meet the basic survival of people in disaster areas as the principle to calculate the needs of food,cold and medical supplies.The timely estimation of the demand for emergency materials after the earthquake can reduce the inaccurate supply of materials to a large extent,so that resources can be saved while ensuring the basic needs of the people in the disaster areas as far as possible.Thus it is not wasted but also contributes to the successful implementation of the rescue work in the later stage.