Comparison of recommendation algorithms for user groups: a food-based case study

Abstract

There is a rise in the development of platforms that work with the distribution of buying and selling food, and with the increase in food options and the number of users, such platforms use recommendation systems to facilitate the user’s choice. These recommendations are usually based on information that the algorithm obtains previously. And increasingly, these recommendations need to be right in specific contexts. This article proposes to compare, through common metrics in the literature, the use of two recommendation algorithms in a context of user groups to make a joint recommendation. One of the algorithms uses a database of groups in neural network training, and the other algorithm uses databases of auxiliary domains with different contexts to perform the prediction. The results indicate that it is possible to perform the prediction for groups of users even if a database with scarce data is used. The article is a theoretical basis to show the efficiency of recommending it to groups in the food domain, and can be incorporated and added to existing platforms.