A Method for Recommending Additional Orders Considering Order Time Distribution in Restaurants

Abstract

In recent years, there are many recommendation systems that recommend products suited to individual users, such as "recommended products" on mail-order websites. Among them, the product recommendation method based on collaborative filtering is well known as a method for recommending products that are related to the products purchased by customers. When considering product recommendation for additional orders using collaborative filtering in restaurants, it is necessary to take into account the trend of order time for each product, since the main dish may be recommended even after a meal because the time factor is not taken into account in the recommended products. In this study, we propose a product recommendation method based on order time distribution by calculating the distribution of the ordering time for each product using kernel density estimation. We compared the proposed method with a conventional collaborative filtering method on actual restaurant data, and confirmed that the proposed method has higher recommendation accuracy than the conventional method.