Forecast of Large Earthquake Emergency Supplies Demand Based on PSO-BP Neural Network

Abstract

Since ancient times, earthquakes have been frequent in China. They have caused serious damage to people's lives and the economic conditions of the country. A large earthquake will cause serious casualties, and a large number of emergency supplies will be needed in the disaster area. However, since earthquakes regularly occur quickly, it is impossible to determine the demand for emergency supplies. Therefore, emergency supplies lose accuracy in distribution. In this paper, we adopt an indirect prediction method. We use particle swarm optimization to improve and optimize the initial weights and thresholds of the BP neural network. Then we predict the mortality rate and injury rate of a large earthquake. Hence the number of casualties and survivors can be obtained. Finally, quantitative relationships between the number of survivors, injured and different supplies are used to estimate the demand for various supplies. By comparing the BP neural network before the improvement, we find that the improved model has higher prediction accuracy and less prediction error. In addition the simulation value fits better with the desired output value. This paper enriches the modeling method for the study of demand prediction of large earthquake emergency supplies.

Keywords

• BP neural network
• demand forecasting
• emergency supplies
• large earthquake
• particle swarm optimization (PSO)