摘要: Accompanied with the development of the electronic commerce, a logistic company should make full use of its inventory resource, transportation resource and some other service resources in order to keep a logistic efficiency and wi... 展开 Accompanied with the development of the electronic commerce, a logistic company should make full use of its inventory resource, transportation resource and some other service resources in order to keep a logistic efficiency and win the competitive advantage. In order to solve the unused resource problem, we introduce the "Logistic Service Overbooking Rate". With the overbooking method, the company can reduce the cost of unused resource and increase the profit. But it also causes the problem that some customers who have booked the service can't be served, and the company should find some methods to compensate the bumped customers. To solve the overbooking problem, we built two models. The first one is a simple statistical model. In the model, we deduce a function of the revenue on average in a short term for a single logistic service. But some actual unfixed factors such as security and customers' fear could not be taken into account because they may have little affect on the current revenue but may affect the further revenue. For this reason we preset another dynamic model based on neural network. Our model is constructed based on Multilayer Feed Forward BP neural network. It contains two processes: training and forecasting. The most important of advantage is that it can forecast future data precisely and can reach optional status automatically. We test the model and the results indicate the feasibility of the model. Furthermore, through the test, we calculate the revenue in several time steps for each method of handling the bumped customers' demand. From the result, different methods can lead to different results for a short term and a long term respectively. 收起