Researchers have developed CaST-POI, a novel approach to next Point-of-Interest (POI) recommendation that improves accuracy by considering the geographic location of candidate POIs. Unlike previous methods that compress user trajectories into a single vector, CaST-POI conditions the user representation on the specific candidate POI being scored. This is achieved by incorporating biases into the attention mechanism that account for the recency and spatial distance of past visits relative to the candidate location. Experiments on datasets from New York City, Tokyo, and California demonstrated significant improvements in recommendation accuracy, with CaST-POI outperforming strong baselines by up to 14.5%. An ablation study indicated that the gains are primarily due to the explicit revisit gate and the candidate-relative spatial bias. AI
IMPACT Enhances location-based recommendation systems by incorporating geographic context, potentially improving user experience in navigation and discovery apps.
RANK_REASON The cluster contains a research paper detailing a new model for POI recommendation. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
