A recommendation system for spots in location-based online social networks

Centralized Online Social Network services (OSN) are collecting immense amounts of data, containing a wealth of information about preferences of their users. Its exploitation for the benefit of the users, even though quite promising, has not seriously been tackled, yet. For this purpose, we propose a personalized recommender for places in location-based OSNs, based on the check-ins of the entire user base. Following a brief analysis, we first propose an interpretation of the data available to OSN providers and an recommendation scheme based on regularized matrix factorization. To evaluate our approach we acquire a large sample of a real data set by crawling Gowalla, one of the most popular location-based OSNs. An exhaustive experimental evaluation then confirms the feasability of using Collaborative Filtering techniques to make personalized recommendation of potentially interesting spots.