Related Experiment Video
Updated: Nov 25, 2025
13:00
Measuring Attention and Visual Processing Speed by Model-based Analysis of Temporal-order Judgments
Published on: January 23, 2017
10.1K
STAN: Spatio-Temporal Attention Network for Pandemic Prediction Using Real World Evidence
Arxiv
|December 17, 2020
Summary
A new spatio-temporal attention network (STAN) improves COVID-19 outbreak predictions by integrating electronic health records and geographical data. This deep learning model offers more accurate and robust pandemic forecasting than traditional methods.
Area of Science:
- Epidemiology
- Deep Learning
- Public Health
Background:
- COVID-19 pandemic necessitates accurate outbreak prediction models.
- Existing epidemiological and deep learning models have limitations in accuracy and robustness.
- Accurate prediction is crucial for effective public health interventions and resource allocation.
Approach:
- Developed a spatio-temporal attention network (STAN) integrating electronic health records (EHR), demographic, and geographical data.
- Utilized an attention-based graph convolutional network to capture spatio-temporal trends.
- Incorporated a physical law-based loss term to enhance long-term prediction accuracy.
Key Points:
- STAN leverages EHR data for local disease status and resource utilization insights.
- Demographic similarity and geographical proximity are considered for improved prediction.
- The model integrates pandemic transmission dynamics into deep learning frameworks.
- STAN demonstrated superior performance over SIR, SEIR, and other deep learning models.
Conclusions:
- STAN provides more accurate pandemic modeling by effectively utilizing patient and geographical data.
- The model captures disease status and medical resource utilization for enhanced predictions.
- Physical law-based regularization contributes to STAN's robust long-term forecasting capabilities.
