Spatial heterogeneity presents significant challenges in improving the spatiotemporal prediction (STP) performance of geographical phenomena. This study identified two primary interpretations of spatial heterogeneity: variable heterogeneity, which refers to spatially uneven distributions of variables, and relational heterogeneity, which involves spatial variations in the relationships between variables. Within STP models, relational heterogeneity is a significant factor that influences performance; however, it is often overlooked by most studies that primarily focus on spatiotemporal dependency or on variable heterogeneity. The dynamics of relational heterogeneity and spatial invariance components require further investigation for predictive learning. Therefore, in this study, we developed a novel Geographical Meta-Learning Neural Network (GeoMetaNet) to address these issues. GeoMetaNet consists of a global component for spatial invariance learning and a local component for dynamic relational heterogeneity learning. In the local component, a meta-learning strategy adjusts the model parameters in different regions with various geographical environments, subject to the effects of spatiotemporal dependency and geographical similarity. We evaluated GeoMetaNet’s performance in predicting cellular traffic in Milan, demonstrating it outperformed several state-of-the-art STP models across 15 STP tasks. We analyze GeoMetaNet’s effectiveness in learning dynamic relational heterogeneity and explore geographical meta-knowledge utility for downstream mining tasks.

Forecasting the spatiotemporal distribution of mobile traffic is crucial for efficient cellular network management. Despite the superior performance of many deep learning studies, they remain inadequate for multi-city forecasting due to the neglect of geospatial effects in deep models. Specifically, spatial heterogeneity and geographical similarity suggest that distinct patterns exist within different urban regions, while shared patterns exist across different cities. To address this gap, this study proposes a Collaborative Specific-Shared Knowledge Learning (CSSKL) framework based on a meta-learning strategy for mobile traffic forecasting in two distinct cities. CSSKL consists of two key components: (1) a geographical learning module for capturing specific patterns using regional customization and (2) a geographical transfer strategy for capturing shared patterns using an attention mechanism. The effectiveness of CSSKL is validated through real-world mobile traffic datasets from two cities, namely Milan and Trentino, Italy. Experimental results demonstrate that CSSKL outperforms all baseline models, yielding a significant improvement in cross-city forecasting performance.

Fine-grained urban traffic data are often incomplete owing to limitations in sensor technology and economic cost. However, data-driven traffic analysis methods in intelligent transportation systems (ITSs) heavily rely on the quality of input data. Thus, accurately estimating missing traffic observations is an essential data engineering task in ITSs. The complexity of underlying node-wise correlation structures and various missing scenarios presents a significant challenge in achieving high-precision estimation. This study proposes a novel multiview neural network termed MVCV-Traffic, equipped with a cross-view learning mechanism, to improve traffic estimation. The contributions of this model can be summarized into two parts: multiview learning and cross-view fusing. For multiview learning, several specialized neural networks are adopted to fit diverse correlation structures from different views. For cross-view fusing, a new information fusion strategy merges multiview messages at both feature and output levels to enhance the learning of joint correlations. Experiments on two real-world datasets demonstrate that the proposed model significantly outperforms existing traffic speed estimation methods for different types and rates of missing data.

The estimated time of arrival (ETA) plays a vital role in intelligent transportation systems and has been widely used as a basic service in ride-hailing platforms. Obtaining a precise ETA is a challenging task due to the complexity of the real-world geographic and traffic environments. Previous works suffer from heterogeneous sparse data learning and multiple-correlation extraction issues. Therefore, this paper presents a hybrid deep learning framework (HSETA) to estimate the vehicle travel time from massive data. First, we encode heterogeneous data to represent various features in different respects. Then, we develop an ensemble factorization machine block (EFMB) structure combined with gated recurrent unit (GRU) and multilayer perceptron (MLP) to extract information from sparse and dense features. Next, the multiple-correlation learning block (MCLB) structure that we propose is utilized to aggregate information based on multiple correlations. Finally, the travel time can be estimated by simple regression. Our extensive evaluations on two real-world datasets show that HSETA significantly outperforms all baselines. Our PyTorch implementation of HSETA and sample data are available athttps://github.com/LouisChenki/HSETA