The field of spatiotemporal prediction is moving towards more accurate and context-aware models. Recent developments have focused on incorporating geographical and temporal weighting into deep learning models to account for non-stationarity in spatial data. This has led to the creation of novel neural network architectures and extensions to existing frameworks, which have shown promising results in improving predictive evaluation scores. Noteworthy papers include:

• Advancing Spatiotemporal Prediction using Artificial Intelligence, which presents mathematical advances to the Geographically and Temporally Weighted Neural Network framework.
• Adaptive Integrated Layered Attention, which proposes a neural network architecture that combines dense skip connections with adaptive feature reuse mechanisms, achieving state-of-the-art performance on several tasks.
• Spatio-temporal Prediction of Fine-Grained Origin-Destination Matrices with Applications in Ridesharing, which introduces a novel prediction model called OD-CED, demonstrating remarkable results in predicting origin-destination demands.
• DRAN: A Distribution and Relation Adaptive Network for Spatio-temporal Forecasting, which proposes a network capable of dynamically adapting to relation and distribution changes over time, outperforming state-of-the-art methods in weather prediction and traffic flows forecasting tasks.