CaST-POI: Candidate-Conditioned Spatiotemporal Modeling for Next POI Recommendation

arXiv cs.AI / 4/25/2026

💬 OpinionIdeas & Deep AnalysisTools & Practical UsageModels & Research

共有:

Key Points

Next POI recommendation in location-based services often ignores that the importance of a user’s past visits depends on which candidate POI is being scored.
The proposed CaST-POI introduces a candidate-conditioned spatiotemporal model that treats each candidate as a query to dynamically attend to user history.
It further adds candidate-relative temporal and spatial biases to better capture fine-grained mobility patterns tied to the relationship between past visits and each candidate.
Experiments on three benchmark datasets show that CaST-POI outperforms existing state-of-the-art approaches, with especially large gains when the candidate pool is large.
The authors provide implementation code via the linked GitHub repository for reproducibility and adoption.

Abstract

Next Point-of-Interest (POI) recommendation plays a crucial role in location-based services by predicting users' future mobility patterns. Existing methods typically compute a single user representation from historical trajectories and use it to score all candidate POIs uniformly. However, this candidate-agnostic paradigm overlooks that the relevance of historical visits inherently depends on which candidate is being evaluated. In this paper, we propose CaST-POI, a candidate-conditioned spatiotemporal model for next POI recommendation. Our key insight is that the same user history should be interpreted differently when evaluating different candidate POIs. CaST-POI employs a candidate-conditioned sequence reader that uses candidates as queries to dynamically attend to user history. In addition, we introduce candidate-relative temporal and spatial biases to capture fine-grained mobility patterns based on the relationships between historical visits and each candidate POI. Extensive experiments on three benchmark datasets demonstrate that CaST-POI consistently outperforms state-of-the-art methods, yielding substantial improvements across multiple evaluation metrics, with particularly strong advantages under large candidate pools. Code is available at https://github.com/YuZhenyuLindy/CaST-POI.git.