A Multi-Channel Next POI Recommendation Framework with Multi-Granularity Check-in Signals

TL;DR

The paper tackles sparsity in next POI recommendation by proposing MCMG, a multi-channel framework that leverages both fine-grained POI signals and coarse region/global signals. It combines a Global User Behavior Encoder (GCN on an all-user POI graph) with a Local Multi-Channel Encoder (location, region, and category channels using self-attention and contextual embeddings) and a Region-aware Weighting Strategy that adaptively fuses channel outputs based on region-patterns. Experimental results on CAL, PHO, SIN, and NY demonstrate that MCMG outperforms state-of-the-art baselines, with notable gains from incorporating region-level signals and the dynamic weighting mechanism. The work provides design insights for robust next POI prediction in sparse data settings and underscores the value of multi-granularity region context in sequential recommender systems.

Abstract

Current study on next POI recommendation mainly explores user sequential transitions with the fine-grained individual-user POI check-in trajectories only, which suffers from the severe check-in data sparsity issue. In fact, coarse-grained signals (i.e., region- and global-level check-ins) in such sparse check-ins would also benefit to augment user preference learning. Specifically, our data analysis unveils that user movement exhibits noticeable patterns w.r.t. the regions of visited POIs. Meanwhile, the global all-user check-ins can help reflect sequential regularities shared by the crowd. We are, therefore, inspired to propose the MCMG: a Multi-Channel next POI recommendation framework with Multi-Granularity signals categorized from two orthogonal perspectives, i.e., fine-coarse grained check-ins at either POI/region level or local/global level. Being equipped with three modules (i.e., global user behavior encoder, local multi-channel encoder, and region-aware weighting strategy), MCMG is capable of capturing both fine- and coarse-grained sequential regularities as well as exploring the dynamic impact of multi-channel by differentiating the region check-in patterns. Extensive experiments on four real-world datasets show that our MCMG significantly outperforms state-of-the-art next POI recommendation approaches.

Figures (10)

• Figure 1: The user check-in signals in multi-granularity from two orthogonal perspectives.
• Figure 2: A running example of mining multi-granularity signals from Bob's check-in records. The blue, red and black lines are the trajectories of Alice, Bob and Cary, respectively; and the white blocks help divide different regions (i.e., region $r_1$ and region $r_2$).
• Figure 3: The regions clustered via k-means for the four cities, where the dots with the same color denote a region; and the red star within each region represent the center of that region, calculated by averaging the geographical positions of all POIs in that region.
• Figure 4: (a-b) trajectory and user distribution w.r.t the number of daily-visited regions; (c) trajectory distribution w.r.t the personalized visiting frequency rank of regions; (d-e) successive cross-region check-in distribution w.r.t different time interval and time slots of a day; (f-g) successive cross-region check-in distribution w.r.t distance and category popularity of various regions; (h) successive cross-region check-in distribution w.r.t the transition of different types of regions.
• Figure 5: The reasons for users visiting infrequently-visited regions or crossing regions.