The Research On The Emergency Logistics Risk Analysis Based On The BP Neural Network

In recent years, all kinds of emergencies frequently take place and have caused catastrophic aftermath and had wide influences, such as Indian Ocean tsunami in 2004, the snow disaster of the south and the big earthquake of Wen Chuan in China in 2008, Influenza A virus subtype H1N1 all over the world in 2009. These emergencies have not only caused enormous loss in the lives and properties, but also affect the stability of the society, even jeopardizing the safety of the whole nation. Needless to say, these emergencies have hindered the development of economy.The emergency logistics is a special logistic action in order to provide the emergency material to deal with incidents. There are two aims that emergency logistics systems want to reach. The first is to dealing with incidents at the best early start-time, the second is to reduce the total cost of dealing with these sudden public incidents as it can. The level of emergency logistics is directly related to the fact that whether the emergencies can be controlled, whether the loss can be made minimum and whether the disasters can be magnified. Therefore, it is an exigent need of the development of the trend to enhance risk management in the process of emergency logistics, which plays a key in the success of emergency logistics operation. Meanwhile, carrying on accurate, quick and comprehensive risk analysis is pivotal to the performance of emergency logistics risk management.This paper carried out a new appraising method which is based on the BP neural network. The method was carried on the base of the research on the appraising index system and appraising method of emergency logistics risk which already exist. Facing to the shortage of the method which exists in the appraising the emergency logistics risk currently, the paper built an appraising index system first. Then according to the index system, designed the BP nerve network model, and gave out the feasible valuation procedure. On reckoning, the paper used the MATLAB neural networks tool case (NNT) to design the net and calculate. Training and testing by studying a sample book, made the model error reach the predetermined range inner. Then an entire network model building-up was finished. Finally, the paper verified the accuracies and the feasibility of this method with the example.