Spatial-Temporal Aware Self Attention Mechanism for Next POI Recommendation

Due to the swift growth and widespread adoption of Internet technology and mobile devices, globalizing information has become an undeniable trend, and next Point-of-Interest (POI) recommendation has been widely used in daily life. Currently next POI recommendation has achieved good results in life, but the existing work still has some problems. Most of methods discretely encode users' spatial-temporal preferences when learning their spatial and temporal preferences, which may lead to loss of information during the discrete encoding process. To tackle the mentioned issue, we introduce STASAN, a Spatial-Temporal Aware Self Attention Mechanism for Next POI Recommendation. Specifically, we design a spatial-temporal aware self-attention layer for capturing users' personalized spatial-temporal preferences. Finally, we conduct experiments on two public datasets, and the results indicate that the STASAN model proposed in this paper outperforms existing state-of-the-art models.