MWED-Former: A Multi-Scale Weather-Aware Transformer for Emergency Department Demand Forecasting with Weather Bias Correction
Fundamental Scientific Reports in Multidisciplinary Areas · 25.06.2026 · 2 (02) · 132-145
DOI: 10.66238/fsrma100 ↗Kurz gefasst
Fachartikel in der Zeitschrift Fundamental Scientific Reports in Multidisciplinary Areas, online erschienen am 25.06.2026. Verfasst von Zhuo Chen und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Emergency Department (ED) overcrowding has become a critical challenge for modern healthcare systems, leading to increased patient waiting times, inefficient resource allocation, and reduced service quality. Accurate forecasting of ED demand is therefore essential for intelligent hospital management and emergency resource optimization. However, existing forecasting approaches primarily rely on historical patient arrival records and often neglect the influence of weather conditions or directly utilize Numerical Weather Prediction (NWP) outputs without correcting forecast biases, limiting their effectiveness under complex meteorological environments. To address these issues, this study proposes MWED-Former (Multi-scale Weather-aware Emergency Demand Transformer), a novel deep learning framework that integrates multi-scale weather bias correction with Transformer-based temporal forecasting. Specifically, a Multi-scale Weather Correction Encoder is designed to capture weather patterns at local, city, and regional scales, while a Weather-aware Fusion Encoder employs cross-attention to model interactions between corrected weather features and historical ED demand. Furthermore, a Dual-Temporal Transformer is introduced to jointly learn short-term fluctuations and long-term periodic trends in patient arrivals. Experimental results on real-world hospital and meteorological datasets demonstrate that MWED-Former achieves an MAE of 7.94, RMSE of 11.87, and MAPE of 7.25, outperforming state-of-the-art baselines such as Informer, Autoformer, and PatchTST. Under extreme weather conditions, the model maintains strong robustness with an MAE of 9.18 and improved classification performance. These results confirm that integrating multi-scale weather bias correction with Transformer-based temporal modeling significantly enhances ED demand forecasting accuracy and reliability for healthcare resource planning.
Zitieren
APA
Chen, Z., Zhu, Q., & Zeng, Y. (2026). MWED-Former: A Multi-Scale Weather-Aware Transformer for Emergency Department Demand Forecasting with Weather Bias Correction. Fundamental Scientific Reports in Multidisciplinary Areas, 2(02), 132-145. https://doi.org/10.66238/fsrma100
Vancouver
Chen Z, Zhu Q, Zeng Y. MWED-Former: A Multi-Scale Weather-Aware Transformer for Emergency Department Demand Forecasting with Weather Bias Correction. Fundamental Scientific Reports in Multidisciplinary Areas. 2026;2(02):132-145. doi:10.66238/fsrma100
BibTeX
@article{Chen2026,
author = {Chen, Zhuo and Zhu, Qianya and Zeng, Yan},
title = {MWED-Former: A Multi-Scale Weather-Aware Transformer for Emergency Department Demand Forecasting with Weather Bias Correction},
journal = {Fundamental Scientific Reports in Multidisciplinary Areas},
year = {2026},
volume = {2},
number = {02},
pages = {132--145},
doi = {10.66238/fsrma100}
}
EUVIMED