Application of the model combining demand forecasting and inventory decision in feature based newsvendor problem

This paper consider a one-step machine learning algorithm and propose a prediction technology using a large amount of relevant information. In the method proposed in this paper, the manager observed P features related to demand data n times, and established a prediction model through scikit-learn machine learning framework to study the data-driven newsvendor problem. In particular, the research method of this paper does not need to obtain the accurate demand distribution before optimization, and the data used by the algorithm is not limited to the historical sales data. We filter multiple characteristic variables related to the demand, substitute the filtered characteristic variable set into the model to predict the future demand, and make the order quantity decision of the newsvendor model according to the predicted demand. Finally, a case study was conduct to verify the effectiveness and feasibility of the proposed method. We find that using multiple characteristic information to predict the future demand has important guiding significance for the order quantity decision of newsvendor model, and this method can be applied to the inventory decision of many commodities.