Deep Spatiotemporal Model for COVID-19 Forecasting.

Area of Science:

• Epidemiology
• Artificial Intelligence
• Public Health

Background:

• COVID-19 has caused significant global mortality.
• Machine learning offers an alternative to conventional models for epidemiological forecasting.
• Existing models often overlook the spatial component of virus spread.

Purpose of the Study:

• To propose a novel deep learning model integrating spatial and temporal analysis for COVID-19 forecasting.
• To enhance the accuracy of short- and medium-term predictions for SARS-CoV-2 spread.
• To support health authorities in optimizing resource allocation and policy implementation.

Main Methods:

• A hybrid deep learning model combining Convolutional Neural Networks (CNN) for spatial analysis and Long-Short Term Memory (LSTM) Recurrent Neural Networks (RNN) for temporal analysis.
• Application of the model to a sequence of COVID-19 incidence images.
• Validation using data from 286 primary care centers in Madrid, Spain.

Main Results:

• The proposed hybrid CNN-LSTM model demonstrated improved performance compared to models focusing solely on temporal patterns.
• Significant enhancements in root mean square error (RMSE) and explained variance (EV) were observed.
• The model effectively captured both spatial and temporal dependencies in COVID-19 spread.

Conclusions:

• The integrated spatial-temporal deep learning approach offers superior COVID-19 forecasting accuracy.
• This model provides a valuable tool for public health decision-making in managing infectious disease outbreaks.
• Future research should explore the scalability and adaptability of this model to other geographical regions and diseases.