STAGformer, a new Spatio-temporal Agent Graph Transformer, was introduced by Ye Zihao on July 7, 2026, to enhance the accuracy of bike-sharing demand forecasting. This innovative model addresses the challenges posed by complex spatio-temporal dependencies and the vast scale of urban networks.
The STAGformer achieves efficient global modeling with linear computational complexity, making it a significant advancement in the field of machine learning. The model’s unique two-step agent attention mechanism allows for the aggregation of global information, which is then effectively broadcast back to individual stations and time steps, capturing long-range interactions while minimizing the computational cost.
Key Features of STAGformer
The STAGformer integrates four core modules:
- Spatio-temporal encoder: This module fuses dynamic node features with external contextual factors such as weather, time, and points of interest.
- Graph propagation module: It facilitates spatial neighbor aggregation to improve accuracy.
- Temporal convolution module: This extracts local patterns, enhancing the model's predictive capabilities.
- Agent attention module: It models global dependencies, proving critical for effective demand forecasting.
Performance Evaluation
Extensive experiments conducted on two real-world datasets, namely NYC Citi-Bike and Chicago Divvy-Bike, reveal that STAGformer consistently surpasses state-of-the-art baselines across various prediction horizons. The model achieved remarkable improvements in both RMSE and MAE metrics, showcasing its efficacy in micro mobility demand forecasting.
Ablation studies further validate the significance of each component, with particular emphasis on the agent attention mechanism’s role in modeling global spatio-temporal dependencies.
Conclusion and Future Implications
The introduction of STAGformer marks a pivotal moment in the realm of micro mobility demand forecasting, providing a robust tool for urban planners and bike-sharing operators. By enhancing the accuracy of station-level demand predictions, STAGformer could revolutionize the efficiency of bike-sharing systems, contributing to smarter urban mobility solutions.
🤖 This article was rewritten by Feed and Figures' editorial AI from a report originally published by arXiv Machine Learning. Facts and quotes are preserved from the original; the rewrite focuses on clarity and structure. For the unedited original, see the source link below.
