A Neural Network Model Based on Quadratic Programming to the Single-period and Multi-product Newsvendor Problem

The newsvendor problem now has been extended into complicated models to meet the commerce pattern’s growing intricacy. For the requirement of outsourcing operation, supply chain management and cost-profit control, we need to develop practicable, potent and computeroperated approaches to resolve the newsvendor problems which the various costs and characteristic parameters of products are divers and the constraints of trade operation are multiplex. In this paper, we develop a quadratic programming approach (QPA) to settle the single-period, multi-product and multi-constraint newsvendor problem with regular distribution of market demand, and then we apply neural network method (NNM) to calculate the optimal order quantity of each potential available product. This model is quite practicable in actual application for that the QPA transforms the original newsvendor problem from integral pattern to quadratic function, which not only increases the maneuverability but also decreases the complexity of calculation greatly, and the NNM establishes a general model which is expansible and superior especially when the number of products is large.