Location recommendation based on periodicity of human activities and location categories

Abstract

Location recommendation is a popular service for location-based social networks. This service suggests unvisited sites to the users based on their visiting history and site information. In this paper, we first present how to build the temporal and spatial probability distribution functions (PDF) to model the temporal and spatial checkin behavior of the users. Then we propose two recommender algorithms, Probabilistic Category Recommender (PCR) and Probabilistic Category-based Location Recommender (PCLR), based on the periodicity of user checkin behavior. PCR uses the temporal PDF to model the periodicity of users' checkin behavior. PCLR combines the temporal category model used in PCR with a geographical influence model built on the spatial PDF. The experimental results show that the proposed methods achieve better precision and recall than two well-known location recommendation methods. © Springer-Verlag 2013.