Successive POI Recommendation via Brain-inspired Spatiotemporal Aware Representation

POI vector representation (embedding) is the core of successive POI recommendation. However, existing approaches only rely on basic discretization and interval analyses and fail to fully exploit complicated spatiotemporal attributes of POIs. Neuroscience research has shown that the mammalian brain entorhinal-hippocampal system provides efficient graph representations for general knowledge. Moreover, entorhinal grid cells present concise spatial representations, while hippocampal place cells represent perception conjunctions effectively. Thus, the entorhinal-hippocampal system provides a novel angle for spatiotemporal aware representation, which inspires us to propose the SpatioTemporal aware Embedding framework (STE) and apply to POIs (STEP). STEP considers two types of POI-specific representations: sequential representation and spatiotemporal conjunctive representation, learned using sparse unlabeled data based on the proposed graph-building policies. Notably, the spatiotemporal conjunctive representation represents POIs from spatial and temporal aspects jointly and precisely. Furthermore, we introduce a user privacy secure successive POI recommendation method using STEP. Experimental results on two datasets demonstrate that STEP captures POI-specific spatiotemporal information more accurately and achieves the state-of-the-art successive POI recommendation performance. Therefore, this work provides a novel solution to spatiotemporal aware representation and paves a new way for spatiotemporal modeling-related tasks.