Enhancing Pandemic Prediction: A Deep Learning Approach Using Transformer Neural Networks and Multi-Source Data...

Area of Science:

• Epidemiology
• Computational Biology
• Artificial Intelligence

Background:

• The Covid-19 pandemic underscored the critical need for precise pandemic trend prediction.
• Traditional forecasting methods often struggle with the complex temporal and spatial dynamics of infectious disease outbreaks.

Purpose of the Study:

• To develop and evaluate a deep learning model for accurate county-level prediction of Covid-19 cases and deaths.
• To assess the impact of multi-source data fusion, including social media sentiment, on forecasting accuracy.

Main Methods:

• Utilized transformer neural networks with multi-source data fusion (historical cases, deaths, social media sentiment).
• Incorporated multi-level and multi-scale attention mechanisms for adaptive time-frequency analysis.
• Evaluated model performance across three Omicron variant waves (December 2021 - February 2023).

Main Results:

• The model demonstrated strong performance in predicting county-level Covid-19 cases and deaths.
• Median county agreement accuracy for one-week case forecasts ranged from 74.0% to 82.6%.
• Median county agreement accuracy for one-week death forecasts ranged from 83.2% to 86.3%.

Conclusions:

• The proposed deep learning model significantly improves upon baseline persistence models for Covid-19 forecasting.
• Integrating real-time data and social media sentiment enhances the capture of complex pandemic dynamics.
• The model's accuracy and generalizability offer a valuable tool for public health preparedness and response to future outbreaks.