LGSA: A next POI prediction method by using local and global interest with spatiotemporal awareness

Predicting the next Point-of-Interest (POI) is a persistent issue in the realm of Location-Based Social Networks (LBSN). To discover the user’s dynamic interests and the dependence between different POIs in user trajectory sequences, self-attention based methods have been applied recently. These methods, however, limit local dependence in the customized spatiotemporal region and do not take the special time period in trajectory sequence into account. To this end, we propose a spatiotemporal awareness model with global and local interest (LGSA) for next POI prediction. According to the geographic distance and time interval, each user’s trajectory sequence is separated into individualized spatiotemporal regions, and the dependence between POIs check-in by user in these regions is learned from the local view. Besides, we use a non-invasive way to fuse the user’s trajectory sequence and the time period of the sequence, and mine the user’s dynamic preferences on the time period from the global view. Extensive experiments on three real-world datasets show that LGSA outperforms state-of-the-art methods.