CityVoyager: an outdoor recommendation system based on user location history

Recommendation systems, which automatically understand user preferences and make recommendations, are now widely used in online shopping. However, so far there have been few attempts of applying them to real-world shopping. In this paper, we propose a novel real-world recommendation system, which makes recommendations of shops based on users’ past location data history. The system uses a newly devised place learning algorithm, which can efficiently find users’ frequented places, complete with their proper names (e.g. “The Ueno Royal Museum”). Users’ frequented shops are used as input to the item-based collaborative filtering algorithm to make recommendations. In addition, we provide a method for further narrowing down shops based on prediction of user movement and geographical conditions of the city. We have evaluated our system at a popular shopping district inside Tokyo, and the results demonstrate the effectiveness of our overall approach.