Forecasting dining times in a full-service Thai hotpot restaurant Using Random Forest Classifier
Pongjit Kongnim, Nagul Cooharojananone, Thira Chavarnakul · 2023
In the foodservice industry, time is a crucial factor that impacts both consumers and management. Machine learning (ML) is increasingly used to improve the quality of services through prediction. In this study, we aim to develop a model for predicting meal duration using Random Forest Classification algorithm. The study uses data from the Point-of-Sale (POS) system of a full-service Thai hotpot restaurant, with a focus on two commercial areas in Bangkok. The variables that we used include the branch, the number of customers, the number of items, the day of the week, and the time of the day. As a result, the overall accuracy of the model was 86% and the F1-score was 0.81. The discussion of the potential use of this approach in connection with the existing system in a restaurant could also be beneficial, aiding the restaurant in planning management more efficiently and gaining a better understanding of consumer behavior. This study will discuss the results of the model along with additional perspectives for future work.