Predicting COVID-19 in China Using Hybrid AI Model.

Area of Science:

• Epidemiology
• Artificial Intelligence
• Computational Biology

Background:

• The COVID-19 pandemic necessitates accurate forecasting for effective control.
• Traditional epidemic models often oversimplify infection dynamics by assuming uniform infection rates.

Purpose of the Study:

• To develop an advanced hybrid artificial intelligence (AI) model for precise COVID-19 prediction.
• To improve upon traditional epidemic models by incorporating variable infection rates and public health interventions.

Main Methods:

• An improved susceptible-infected (ISI) model was developed to account for varying infection rates.
• A hybrid AI model was constructed by integrating Natural Language Processing (NLP) and Long Short-Term Memory (LSTM) networks into the ISI model.
• The model was trained and validated using COVID-19 epidemic data from various Chinese provinces and cities.

Main Results:

• The study identified a higher infection rate in individuals within 3-8 days post-infection, aligning with actual transmission patterns.
• The hybrid AI model demonstrated significantly reduced prediction errors compared to traditional models.
• Achieved low Mean Absolute Percentage Errors (MAPEs) of 0.52% (Wuhan), 0.38% (Beijing), 0.05% (Shanghai), and 0.86% (countrywide) for 6-day forecasts.

Conclusions:

• The proposed hybrid AI model offers a more accurate and reliable approach to COVID-19 prediction.
• Incorporating variable infection rates and NLP/LSTM enhances the understanding of epidemic transmission dynamics.
• This model provides a valuable tool for public health strategies in managing and predicting infectious disease outbreaks.