Application of data-based prediction methods in newsvendor problems subject to purchase price uncertainty

Poor procurement decisions, especially involving perishable or short life-cycled products, which will have to be disposed of, can cost companies large portions of their profits. The newsvendor problem addresses inventory decisions to assist retailers in deciding just the right order quantity while still subject to uncertainty. Efficient time series forecasting techniques, including the use of machine learning models, have helped reduce uncertainty and improve financial results by offering insight on future outcome-based decisions. This work proposes a hybrid model, exploring linear and nonlinear modeling capabilities of classic, modern and machine learning models which are applied to a financial time series in order to anticipate fluctuations in the price of supplies purchased by the newsvendor. This information is used to aid the decision-maker in an optimization problem involving both the decision on the order quantity and the best moment for the one-time per cycle newsvendor replenishment in situations where the purchase price fluctuates in time. The proposed model outperformed both SARIMA and Prophet models as efficient pointers to the best moment to place an order. A case study of a Brazilian supermarket chain which must place weekly orders of perishable goods, specifically of meat products, was chosen to illustrate the methodology.