Forecasting Daily Visitors and Menu Demands in an Indonesian Chain Restaurant using Support Vector Regression Machine

Abstract

Demand fluctuation is a critical factor in the everyday operating choices made by a restaurant. The aim of this study is to investigate menu demand forecasting in restaurants using Multiple Regression and Support Vector Regression Machine (SVR) algorithms to forecast potential visitors and menu demand using point-of-sale (POS) data. A model for predicting store-specific demand is proposed that takes into account variables such as seasonality, public holidays, and order peak times. The model's verification using fundamental restaurant data demonstrates that SVR will produce a percentage error of as low as 14.84 percent when forecasting restaurant guests and 31.2 percent when predicting restaurant menu demand. The results demonstrate that this approach is practical for forecasting revenue and consumer counts, as well as demonstrating that managers will learn about the variables that influence customer behaviors. There are extensive discussions and suggestions for potential studies on predicting and planning management in chain restaurant operations.