A restaurant recommender system for a new-born app-based gastronomic guide

Abstract

[en] This is a project of applied data analysis made in collaboration with the CPO of the restaurants app Velada. We analyze the data collected by the app with the objective of building a recommender system for its restaurants. The initial objective was to give particular attention to the parameter time, building a model able to make the right recommendation in the right moment for each user of the app. Different attempts have been performed, using both collaborative filtering and content based recommender systems, with different types of information as data input. We show that the best approach is a binary collaborative filter, although the results are preliminary due to the lack of enough data. We also show that the performance will be easily improved as new data becomes available. Finally, we provide some insights on how the problem of making a recommendation for a restaurant just in time could be solved in the future.