Next point-of-interest recommendation

Abstract

In the recent years, Next Point-of-Interest (POI) recommendation system has become more popular. The goal of POI recommendation system is to give POI recommendation to users given the users' historical check-in history. It is important to take into account the users recent check-in sequence and their preference to give accurate recommendations. This report proposes a POI recommendation model that utilizes multi-task learning that considers both the long-term preference and short-term preference of the users. The long-term component will learn about the user preference, and the short-term component will learn about the recent sequential check-in. The performance of the proposed model will then be compared to other baseline models to highlight the advantages of the proposed model.