A Multi-Task Embedding Based Personalized POI Recommendation Method
Chen Ling,Ying Yuankai,Lyu Dandan,Yu Shanshan,Chen Gencai
DOI: https://doi.org/10.1007/s42486-021-00069-z
2021-01-01
CCF Transactions on Pervasive Computing and Interaction
Abstract:Point of interest (POI) recommendation can help people explore new regions. Existing methods tend to exploit different information to improve recommendation performance. However, there are still challenges in how to integrate different information flexibly and effectively. Therefore, we propose a multi-task embedding based personalized POI recommendation method (MTEPR). On one hand, multi-information (sequential, social, temporal, geographical, semantic, gender, and preference information) are exploited to embed users and POIs. On the other hand, MTEPR incorporates sequence and graph embedding methods to model user location history, as well as to preserve the relationships between users, POIs, and different information. Specifically, sequential information is captured by applying a sequence embedding method on users’ check-in sequences, and other information are captured by applying graph embedding methods on User–User, User-Time Slot, POI-Time Slot, POI-Region Hierarchy, POI-Category Hierarchy, User-Gender, and User-POI graphs. The scores of candidate POIs are calculated based on the embeddings to make recommendations. We conduct extensive experiments to evaluate the performance of MTEPR on two real-world datasets, and the results show its superiority over baselines. We also study the impact of different information, and the results reveal that temporal and semantic information contribute most to improving recommendation performance.