Cross-sectional analysis and data-driven forecasting of confirmed COVID-19 cases

Area of Science:

• Epidemiology and Public Health
• Computational Biology and Bioinformatics
• Data Science and Machine Learning

Background:

• Coronavirus disease 2019 (COVID-19) is a major global mortality cause.
• Previous models for COVID-19 spread lack comprehensive data, limiting understanding.
• Regional attributes like geographic and local factors are crucial but often overlooked.

Purpose of the Study:

• To develop a data-driven model for comprehensive cross-sectional analysis and forecasting of the COVID-19 pandemic.
• To identify critical features influencing COVID-19 infection rates.
• To improve the accuracy of confirmed COVID-19 case forecasting.

Main Methods:

• Utilized daily historical time-series data and regional attributes for modeling.
• Employed XGB (eXtreme Gradient Boosting) with SHAP (SHapley Additive exPlanation) to determine critical features influencing infection rates.
• Applied a Dual-Stage Attention-Based Recurrent Neural Network (DA-RNN) for accurate case forecasting, comparing it with Support Vector Regression (SVR) and encoder-decoder networks.

Main Results:

• Identified critical features significantly influencing COVID-19 infection rates using XGB and SHAP.
• The DA-RNN model demonstrated superior performance over SVR and encoder-decoder networks in forecasting confirmed COVID-19 cases.
• Model performance was rigorously evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared (R²) metrics.

Conclusions:

• Integrating regional factors with time-series data significantly enhances COVID-19 pandemic modeling.
• The DA-RNN model offers a more accurate approach to forecasting confirmed COVID-19 cases.
• This study provides valuable references for effective COVID-19 pandemic control and prevention strategies.