Exploring Temporal and Spatial Features for Next POI Recommendation in LBSNs

With the increasing popularity of Location-Based Social Networks (LBSNs), a significant volume of check-in data of users has been generated. Such massive data brings difficulties for the users to efficiently retrieve their desired point-of-interest (POI). As a result, POI recommendation systems have received extensive attention from academia and industry. Currently, most existing POI recommendation approaches only provide users with a fixed set of recommended POIs based on the historical check-in records of the users, and cannot achieve flexible and feasible recommendations according to different spatial and temporal situations of the users. In this paper, we propose a next POI recommendation model that will predict POIs to be visited by users in the next few hours according to their historical check-in data and current contextual information (such as the current time and locations of the users). In our model, we propose a unified approach to calculate context-aware similarities between different users by investigating the influences of both temporal and spatial features for the users. We also propose an approach to dynamically generate different POI recommendation lists for a particular user according to different current context information of the user. Compared with the state-of-the-art POI recommendation approaches, the experimental results demonstrate that our system achieves much better performance.