A System Prototype for Food Sales Forecasting and Optimization to Reduce Food Waste for Short-Shelf-Life Products

Abstract

The reduction of food waste is one of the major challenges of today for retailers and wholesalers. Large amounts of food are thrown away on the retail and wholesale level per year. Since globally available resources are limited, preventing food waste is a very important way to reduce the carbon footprint and even help protect the environment because the production of goods consumes both large amounts of energy and land. Preventing food waste is intertwined with the related problem of order generation. The generation of orders depends on accurate forecasts provided to the users. In this paper, we present a system description of a prototype that signicantly improves forecasts to facilitate the reduction of food waste through the use of machine learning to provide a basis for subsequent order optimization. Our system has been developed in cooperation with Austrian retailers and wholesalers who provide both real-world data and valuable insights into the inner workings of Austrian grocers. We present an overview of the system and the technologies utilized to achieve our goals. In addition, we also discuss the constraints and ethical considerations encountered. Our evaluation shows that our system can help achieve the goals of reducing food waste while being very useful to our project partners and, therefore, workable in the real world.