Next Points of Interest Recommendations Based on Spatio-Temporal-Category Pattern Information

To address the issue of insufficiently mining spatio-temporal category transition pattern information in existing methods, this paper proposes a next point-of-interest recommendations model based on spatio-temporal category check-in pattern information. Firstly, the model thoroughly extracts high-level category information and integrates it with the spatio-temporal background information of check-ins to mine transition pattern information. Secondly, gated recurrent units effectively capture long-term dependencies to enhance preference feature representation. Finally, this paper extracts and aggregates subsequence information. The proposed model is evaluated and analyzed on the Foursquare and Gowalla datasets, demonstrating its outstanding recommendation performance.