A protocol for meal attendance forecasting at Worcester Academy

In school dining halls, inaccurate meal attendance forecast is the key rationale that leads to either food waste or inefficiencies in which the dining hall runs out of food. The dining service group in Worcester Academy (WA), an independent high school in Massachusetts, United States, has been using some basic forecasting methods required by the contractor company. But it lacks accuracy as the models are not specifically designed for WA, said by the manager. Thus, this study takes a lens into a range of forecasting techniques employed in high schools, universities, and restaurants throughout the world. The techniques include basic time-series models like Naive, Moving Averages, and simple exponential smoothing, to Linear Regression, specialized in ARIMA and Facebook’s Prophet, and machine learning algorithms including Lasso regression. The types of data they collected are also identified. Then, the study proposes a new forecasting protocol for WA based on the comparative analysis of methods. This protocol aims for finding the best methods of forecasting at WA. This study will be applied or at least referenced in WA dining service during school years in the future. The best forecasting methods will hopefully be implemented continuously to improve dining hall operations, if feasible.