Forecasting the Number of Customers Visiting Restaurants Using Machine Learning and Statistical Method

Tanizaki, T.; Kozuma, S.; Shimmura, T.

In this paper, it is proposed the forecasting of the number of customers visiting restaurants using machine learning and statistical method. There are some researches on forecasting the number of customers visiting restaurants. Since the beginning of last year, the number of customers visiting restaurants has plummeted due to COVID-19. A machine learning-based approach can be applied to forecast something including stable trends. Therefore, in this paper, machine learning that incorporates the moving average method is proposed to reflect the latest fluctuation trend. Furthermore, a forecasting method using deep learning is proposed to improve forecasting accuracy. In the method using deep learning, the analysis results on the normalization of training data and the contribution of meteorological data to the forecasting accuracy are described. It was found that the introduction of the moving average into the explanatory variables is effective when the trend of the number of customers visiting fluctuates rapidly. It was also found that normalization for each year of training data is effective when the annual average number of customers visiting restaurants increases or decreases monotonically. © 2021, IFIP International Federation for Information Processing.