COVID-19 mortality prediction using ensemble learning and grey wolf optimization.

Area of Science:

• Computational biology
• Medical informatics
• Machine learning in healthcare

Background:

• COVID-19 presents a spectrum of severity, from mild to fatal.
• Accurate risk stratification is essential for optimal patient management and treatment decisions.
• Identifying individuals at high risk for severe outcomes is a critical clinical need.

Purpose of the Study:

• To develop and validate a computational method for estimating COVID-19 patient mortality risk.
• To create a predictive model utilizing ensemble learning and genetic algorithms.
• To identify key features for accurate and efficient risk assessment.

Main Methods:

• Development of an ensemble learning model combined with a genetic algorithm.
• Training and validation using a dataset of 4,711 confirmed SARS-CoV-2 infections.
• Evaluation of model performance using AUCROC (Area Under the Receiver Operating Characteristic Curve).

Main Results:

• The best-performing ensemble model achieved an AUCROC of 0.7802.
• The optimal model requires only 10 features, simplifying data collection.
• Demonstrated robustness and efficiency of the machine learning and genetic algorithm combination.

Conclusions:

• The proposed computational method effectively estimates COVID-19 mortality risk.
• The model's efficiency (fewer features) can reduce diagnostic costs and improve prognostic timelines.
• Machine learning combined with genetic algorithms offers a powerful approach for developing clinical predictive models.