Forecasting the Spread of COVID-19 Using Deep Learning and Big Data Analytics Methods.

Area of Science:

• Epidemiology
• Data Science
• Public Health

Background:

• The COVID-19 pandemic necessitates advanced forecasting strategies utilizing current technology.
• Existing research often focuses on single or multiple countries, lacking comprehensive African regional analysis.
• This study addresses the need for continent-wide investigations into COVID-19 spread.

Purpose of the Study:

• To forecast COVID-19 cases across all five major African regions.
• To identify the most critical countries impacted by the pandemic.
• To compare the efficacy of statistical and deep learning models for epidemiological forecasting.

Main Methods:

• Utilized a univariate time series approach with confirmed cumulative COVID-19 cases.
• Employed statistical models like Autoregressive Integrated Moving Average (ARIMA) and deep learning models including Long Short-Term Memory (LSTM) and Prophet.
• Evaluated model performance using seven metrics: MSE, RMSE, MAPE, sMAPE, PSNR, NRMSE, and R2 score.

Main Results:

• The Long Short-Term Memory (LSTM) model demonstrated superior performance compared to ARIMA and Prophet.
• Mali (Western Africa), Angola (Southern Africa), Egypt (Northern Africa), Somalia (Eastern Africa), and Gabon (Central Africa) were identified as most vulnerable.
• These countries are projected to experience the highest increase in cumulative positive COVID-19 cases: 22.77%, 18.97%, 11.83%, 10.72%, and 2.81%, respectively.

Conclusions:

• Deep learning models, particularly LSTM, offer robust capabilities for forecasting infectious disease spread in Africa.
• Identifying high-risk countries enables targeted public health interventions and resource allocation.
• Comprehensive regional analysis is crucial for effective pandemic containment strategies across the African continent.