A data-driven approach to reduce food waste for a consumer goods company

Official URL: https://risc01.sabanciuniv.edu/record=b2579100_(Table of contents)

Today, the prevention of food waste has become a very significant issue for a sustainable future. In this study, an inventory planning process that will minimize both inventories and lost sales costs and indirectly food waste was studied by analyzing the sales data of a perishable product whose demand is random. The newsvendor problem has been adopted because it is a widely used perishable inventory management problem where the demand is uncertain. The traditional newsvendor problem is implemented on the assumption that the demand distribution is known. However, in reality the true demand distribution is unknown. Therefore, a data-driven and integrated solution method is used in our study by using machine learning models and quantile regression methods that do not require demand distribution knowledge. In the study where we use traditional demand forecasting methods and sequential demand estimation and optimization for comparison, we find that both the integrated demand estimation and optimization methods and machine learning methods perform better than their counterparts.