Next Point of Interest Prediction Using Graph Attention Networks and Cognitive Navigational Styles

Next POI of interest prediction (Next POI) is a fundamental task in both location-based services and recommendation systems. However, previous approaches rarely consider individual differences in navigation strategy and spatial cognition. This paper proposes a new framework that embeds personalized navigational style into a deep learning model based on graph-attention networks. By integrating user-specific spatial preferences into graph representations and model training, the proposed approach dynamically integrates such data from cognitive assessments, including the Spatial Cognitive Style Test and the QSOF questionnaire. The model can therefore adapt to user-specific navigation strategies and significantly improve precision, leading to higher user satisfaction. Our results are testimony to the effectiveness of integration among cognitive insights with advanced graph-based learning to develop human-centered Next POI prediction systems.