CTKGRec: A context-aware temporal knowledge graph reasoning model for next POI recommendation

The next point-of-interest (POI) recommendation task aims to predict the next place of interest for a user based on their historical check-in information and current status, which is vital for improving the quality of location-based services (LBS). Existing methods often leverage rich information to construct various graphs or knowledge graphs that capture higher-order information about users and POIs. However, most knowledge graph-based POI models adopt traditional static knowledge graph methods, which makes it difficult to capture the complex temporal dynamics in user check-in data. That is, the user's check-in data exhibits a certain habitual repetition pattern and is influenced by the global popularity trend and transient factors. In this paper, we propose CTKGRec, a model based on a context-aware temporal knowledge graph that we construct, namely CTKG. The model incorporates user personal preferences and global POI transfer preferences to recommend the next POI. The model employs a context-aware copy mechanism and generates the habit prediction for repetitive check-ins and the novelty prediction for unfamiliar POIs based on the user's memory. Extensive experiments on two large-scale real-world datasets demonstrate that our proposed method achieves state-of-the-art performance.