Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TransTARec: Time-Adaptive Translating Embedding Model for Next POI Recommendation

Yiping Sun

TL;DR

This work tackles next POI recommendation by introducing TransTARec, a time-adaptive translating embedding model that unifies temporal influence, sequential dynamics, and user preference. A neural fusion mechanism produces a time-aware translation vector v_{u,t} from (t_i, u, t_j), enabling the translation v_{p_i} + v_{u,t} ≈ v_{p_j} with a TransH-style hyperplane projection to align embeddings. The model is trained with a margin-based ranking loss and soft constraints to handle large-scale data sparsity, followed by a ranking function that scores candidate POIs for each user and time pair. Empirical results on Foursquare and Mobile datasets show consistent improvements in Top@k metrics over strong baselines, confirming the value of explicitly modeling temporal influence within a translation-based recommendation framework. The approach offers scalable time-aware POI recommendations and lays groundwork for extending to richer knowledge sources.

Abstract

The rapid growth of location acquisition technologies makes Point-of-Interest(POI) recommendation possible due to redundant user check-in records. In this paper, we focus on next POI recommendation in which next POI is based on previous POI. We observe that time plays an important role in next POI recommendation but is neglected in the recent proposed translating embedding methods. To tackle this shortage, we propose a time-adaptive translating embedding model (TransTARec) for next POI recommendation that naturally incorporates temporal influence, sequential dynamics, and user preference within a single component. Methodologically, we treat a (previous timestamp, user, next timestamp) triplet as a union translation vector and develop a neural-based fusion operation to fuse user preference and temporal influence. The superiority of TransTARec, which is confirmed by extensive experiments on real-world datasets, comes from not only the introduction of temporal influence but also the direct unification with user preference and sequential dynamics.

TransTARec: Time-Adaptive Translating Embedding Model for Next POI Recommendation

TL;DR

This work tackles next POI recommendation by introducing TransTARec, a time-adaptive translating embedding model that unifies temporal influence, sequential dynamics, and user preference. A neural fusion mechanism produces a time-aware translation vector v_{u,t} from (t_i, u, t_j), enabling the translation v_{p_i} + v_{u,t} ≈ v_{p_j} with a TransH-style hyperplane projection to align embeddings. The model is trained with a margin-based ranking loss and soft constraints to handle large-scale data sparsity, followed by a ranking function that scores candidate POIs for each user and time pair. Empirical results on Foursquare and Mobile datasets show consistent improvements in Top@k metrics over strong baselines, confirming the value of explicitly modeling temporal influence within a translation-based recommendation framework. The approach offers scalable time-aware POI recommendations and lays groundwork for extending to richer knowledge sources.

Abstract

The rapid growth of location acquisition technologies makes Point-of-Interest(POI) recommendation possible due to redundant user check-in records. In this paper, we focus on next POI recommendation in which next POI is based on previous POI. We observe that time plays an important role in next POI recommendation but is neglected in the recent proposed translating embedding methods. To tackle this shortage, we propose a time-adaptive translating embedding model (TransTARec) for next POI recommendation that naturally incorporates temporal influence, sequential dynamics, and user preference within a single component. Methodologically, we treat a (previous timestamp, user, next timestamp) triplet as a union translation vector and develop a neural-based fusion operation to fuse user preference and temporal influence. The superiority of TransTARec, which is confirmed by extensive experiments on real-world datasets, comes from not only the introduction of temporal influence but also the direct unification with user preference and sequential dynamics.
Paper Structure (11 sections, 13 equations, 6 figures)

This paper contains 11 sections, 13 equations, 6 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Problem Formulation
  5. The Proposed Model
  6. Optimization and Learning the parameters
  7. Ranking POIs for Recommendation
  8. Experiment
  9. Experimental Settings
  10. Performance Comparison and Analysis
  11. Conclusion

Figures (6)

  • Figure 1: Toy example of user check-in sequences.
  • Figure 2: TransTARec Model: the embedding of previous POI $p_i$ is translated to the embedding of next POI $p_j$ via time-adaptive translation vector $\mathbf{v}_{u,t}$.
  • Figure 3: Dataset Analysis
  • Figure 4: Result of Next POI Recommendation on Foursquare and Mobile Dataset
  • Figure 5: Time-specific POI Recommendation on Mobile$_{\leq100}$ Dataset
  • ...and 1 more figures