Research On Prediction Of Demand And Allocation Of Relief Supplies

In recent years, earthquakes have occurred frequently and have brought about great loss to the society. The earthquakes cause huge casualties in a short time. At the same time, a lot of relief supplies which are the basic relief work and basic security are inevitably required. Therefore, the primary task of rescue work is to deliver the supplies needed to the disaster areas as soon as possible. Thus the allocation of relief supplies has an important position throughout the rescue process. At present, research on relief supplies has made considerable progress, but there still exist shortcomings. Based on the previous scholars’study, the paper will do major work on prediction of demand and allocation of relief supplies as follows:Seven evaluating incices were established through the analysis of the main reasons affecting earthquake casualties. In this paper, a BP neural network model for estimating the casualties in earthquakes is presented, in which seven indices values as input layer and the injury rate, mortality rate as the output layer. Then, the method is validated by predicting the demand of Beichuan which was one of the hardest areas in "Wenchuan earthquake".Most of the current researches neglect the potential demand in the time of deliver relief supplies. We proposed a demand predicting model which is based on the classical theory of safety stock and the number of casualties by analogy with the theory of safety stock. A new method is presented to calculate the urgent degree of the disaster areas based on the BP neural network weight contribution rates.A new model for allocation of relief supplies is proposed. In this paper, a bilevel programming model for allocation of relief supplies is constructed. The objective of the upper level is to minimize the time and the lower level is to maximize the relief supplies satisfaction. Finally, the paper takes the "Wenchuan earthquake" in2008as an example to illustrate the operability and the practicality of the model.