Modeling POI-Specific Spatial-Temporal Context for Point-of-Interest Recommendation

Point-of-Interest (POI) recommendation is a fundamental task in location-based social networks. Different from traditional item recommendation, POI recommendation is highly context-dependent: (1) geographical influence, e.g., users prefer to visit POIs that are not far away; (2) time-sensitivity, e.g., restaurants are preferred in dinner time; (3) dependency in a user’s check-in sequence, e.g., POIs planned in a trip. Yet, existing methods either partially leverage such context information or combine different types of contexts using a global weighting scheme, failing to capture the phenomenon that the importance of each context is also context-dependent rather than the same for all recommendation. In this paper, we propose a model to exploit spatial-temporal contexts in a POI-guided attention mechanism for POI recommendation. Such an attention mechanism offers us high flexibility to capture the POI-specific importance of each context. Experimental results on two real-world datasets collected from Foursquare and Gowalla demonstrate that the POI-specific context importance significantly improves the performance of POI recommendation.