Human activity intensity prediction is crucial to many location-based services. Despite tremendous progress in modeling dynamics of human activity, most existing methods overlook physical constraints of spatial interaction, leading to uninterpretable spatial correlations and over-smoothing phenomenon. To address these limitations, this work proposes a physics-informed deep learning framework, namely Gravity-informed Spatiotemporal Transformer (Gravityformer) by integrating the universal law of gravitation to refine transformer attention. Specifically, it (1) estimates two spatially explicit mass parameters based on spatiotemporal embedding feature, (2) models the spatial interaction in end-to-end neural network using proposed adaptive gravity model to learn the physical constraint, and (3) utilizes the learned spatial interaction to guide and mitigate the over-smoothing phenomenon in transformer attention. Moreover, a parallel spatiotemporal graph convolution transformer is proposed for achieving a balance between coupled spatial and temporal learning. Systematic experiments on six real-world large-scale activity datasets demonstrate the quantitative and qualitative superiority of our model over state-of-the-art benchmarks. Additionally, the learned gravity attention matrix can be not only disentangled and interpreted based on geographical laws, but also improved the generalization in zero-shot cross-region inference. This work provides a novel insight into integrating physical laws with deep learning for spatiotemporal prediction.

Cross-domain sequential recommendation (CDSR) utilizes data from multiple domains to recommend the user’s next interaction based on his latest interaction sequence. Currently, many cross-domain sequential recommendation algorithms have been proven to achieve good recommendation performance. However, these algorithms overlook the influence of users’ long-term behavioral patterns and general interests when extracting their current preferences. In this article, we propose a Hierarchical Gating Network for Cross-Domain Sequential Recommendation (HGNCDSR). Specifically, we simultaneously train single-domain and cross-domain interaction sequences, utilizing a hierarchical gating network to capture user interest representations in single-domain and cross-domain, respectively. A feature gating and an instance gating are applied respectively to extract user interests at item feature level and instance level. While learning current preferences from behavior sequences, user representations that reflect behavioral patterns and general interests are simultaneously learned and strengthened. Additionally, we employ the item–item product to model the relationships between candidate items and those in the interaction sequence. Both current interests and item relevance are considered simultaneously, integrating single-domain and cross-domain user preferences to predict the user’s next interaction. We design extensive experiments to show that HGNCDSR has better recommendation performance than other state-of-the-art models.

Digital elevation models (DEMs) are crucial geographical data source whereas the resolution of commonly used DEM products is low and cannot meet requirement of some detailed geo-related applications. Deep learning-based methods have been demonstrated to be effective in super-resolution (SR) techniques, which reconstruct high-resolution (HR) images from low-resolution (LR) images. However, existing deep learning methods have not fully considered the multi-scale spatial heterogeneity and topographic knowledge of DEM data that differentiate them from traditional images. These inevitably lead to the localized smoothing of the reconstructed DEM and influence the reliability downstream geographical analysis. This study proposes a transformer-based topography neural network (TTSR) for DEM SR incorporating a local-global deformable block (LGDB) for capturing the multi-scale spatial heterogeneity and topographic knowledge, a spatio-channel coupled channel attention (SimAM) mechanism for reallocating channel weights and providing a supplement of the global spatial features, and an improved terrain loss function (iLoss) for mitigating noise across datasets. TTSR was validated using two publicly available real-world DEM datasets for recovering DEM from 30 m to 10 m. The root mean square error (RMSE) of the proposed method was reduced by approximately 6-30%, 4-16%, and 1-9% in elevation accuracy, slope accuracy, and aspect accuracy, respectively, compared to the best one of those state-of-the-art methods. This research provides new insights for improving the accuracy of the DEM SR, which will help generate global high-resolution terrain products to geographical studies in the future.

Traffic flow forecasting is crucial to Intelligent Transportation Systems (ITS), particularly for route planning and traffic management. Spatiotemporal graph neural networks have been widely used for this purpose, where a global graph structure is mainly used, but local graph features are ignored. Multi-scale spatiotemporal features—the combined application of global and local graph features—can help effectively discover the underlying spatiotemporal patterns. This study proposed a multi-scale adaptive spatiotemporal forecasting model called Adaptive SpatioTemporal InceptionNet (AST-InceptionNet). To enrich the captured features and extract multi-scale spatiotemporal ones, the inception part is applied to combine local spatiotemporal features with several global ones. Additionally, we developed a fully adaptive graph convolution method, including topologically adaptive graph convolution and an adaptive adjacency matrix. It can autonomously and dynamically learn spatial heterogeneity and enable the proposed model to work with unknown adjacency relations. The proficiency of AST-InceptionNet was validated using four publicly available real-world traffic flow datasets. The experiments demonstrate a satisfactory performance from the proposed model, outperforming existing state-of-the-art methods by up to approximately 12%.