Personalized Interest Sustainability Modeling for Sequential POI Recommendation

Zewen Long, Liang Wang, Qiang Liu, Shu Wu

Abstract: Sequential point-of-interest (POI) recommendation endeavors to capture users' dynamic interests based on their historical check-ins, subsequently predicting the next POIs that they are most likely to visit.Existing methods conventionally capture users' personalized dynamic interests from their chronological sequences of visited POIs. However, these methods fail to explicitly consider personalized interest sustainability, which means whether each user's interest in specific POIs will sustain beyond the training time. In this work, we propose a personalized INterest Sustainability modeling framework for sequential POI REcommendation, INSPIRE for brevity. Different from existing methods that directly recommend next POIs through users' historical trajectories, our proposed INSPIRE focuses on users' personalized interest sustainability. Specifically, we first develop a new task to predict whether each user will visit the POIs in the recent period of the training time. Afterwards, to remedy the sparsity issue of users' check-in history, we propose to augment users' check-in history in three ways: geographical, intrinsic, and extrinsic schemes. Extensive experiments are conducted on two real-world datasets and results show that INSPIRE outperforms existing next POI solutions.