A Sequence Point of Interest Recommendation Algorithm Incorporating Global Features

With the rapid development of location-based social networks and personalized recommendation systems, point-of-interest (POI) recommendation has become increasingly important. However, most existing recommendation methods have failed to fully consider the global information in check-in data and the sequential behavioral patterns in user check-in sequences. To address this issue, this study proposes a POI recommendation algorithm that integrates global features from check-in data and local features from user check-in sequences. The algorithm utilizes graph neural networks to extract global features from check-in data, capturing global check-in patterns and correlations between users. Simultaneously, it processes user historical check-in data through location embedding and bidirectional Transformer encoders to effectively capture sequential preferences in user check-in behavior. Experimental results on two real datasets demonstrate that our approach excels in POI recommendation performance, validating the effectiveness of the algorithm.