Restaurant Demand Forecasting Using Dynamic Neural Network for Business Networking Application

Restaurant is a business establishment that serves public customers with food and drinks. Despite the variety of types of restaurants in the world in general and the Middle East in particular, there are still some common problems that may exist in terms of safety, health and delivery. One recent practically important problem is the disposal of huge amounts of food waste due to the mismatch between the invested demand every month with that purchased by customers. Therefore, estimating the primary demand of various branches of a restaurant company will have a significant saving effect for the head office of that company in addition to the healthy consumption of food as needed for cleaner environment. This study investigates the development of an accurate forecaster of a multi-outlet restaurant via machine learning (ML) techniques. Nonlinear autoregressive network with exogenous inputs (NARX) neural network has been applied to attain reasonably accurate forecasting via real data for a restaurant company of four branches. The forecasting is preferably selected to be in the time-range classification of medium-term forecasting, which takes two months in advance and predicts the next month. The model trained with 70% of the data samples, 15% for verification, and the remainder of 15% for testing to ensure the integrity and robustness of the developed model that closely follows the trends of the data. Quantified analysis of the results has been also reported.