Next point-of-interest (POI) recommendation plays a crucial role in enhancing user travel experiences and driving platform revenues by suggesting potentially appealing locations to users. Existing works have focused on capturing users’ general preferences and dynamic interests by modeling long-term and short-term check-in sequences. However, current long-term models struggle to accurately capture periodic user behaviors, while short-term models often fail to account for users’ personalized geographical preferences within their trajectories. To address these limitations, we propose a novel model: Personalized Short-term and Periodic Long-term Preferences Modeling Network (PSPL). This model integrates users’ short-term spatio-temporal preferences and their long-term periodic location preferences. Specifically, we introduce a S \(^2\) Graph (Spatial Span Graph) used for GCN to model users’ short-term personalized spatial span preferences and devise an ASL Block (Self-Attention and Span LSTM) to capture spatio-temporal preferences and sequential information. Additionally, we employ a Discrete Fourier Transform (DFT)-based method to effectively capture long-term periodic patterns. The integration of these two types of features significantly enhances the accuracy of next POI recommendations. Extensive experiments on real-world datasets demonstrate the superiority of our model, achieving an average improvement of 6.39% in Recall and 6.54% in NDCG compared to the state-of-the-art methods.