Leveraging Machine Learning to Predict Food Waste Quantity: Focusing on Military Dining Facilities as Large‐Scale Food Service Operations

Food waste reveals inefficiencies in resource use and contributes significantly to environmental pollution. In the United States, food waste accounts for the majority of total waste, with more than 50 million tons generated annually. In military units that operate large‐scale institutional food‐service facilities and generate significant food waste, effective protocols for food waste control and management are crucial. Due to the repetitive nature of mass meal services, substantial food waste is consistently generated. Food waste accounts for the largest share of total waste generated across U.S. Army facilities, representing over 50% of total output at some installations. Dining facilities serve as the primary source of this substantial volume. This study develops a machine learning model that accounts for both internal and external environmental factors to predict the quantity of plate waste in military units. The external variables used to identify key factors influencing plate waste generation include weather, date, day of the week, and holidays; meal composition was the internal variable. By ensuring efficient resource use without waste, this study offers insights into achieving environmental sustainability. Finally, the study explores the suitability of different predictive models by comparing the performance of random forest and deep learning approaches and assessing the characteristics of the most effective models.