An Improved Machine-Learning Approach for COVID-19 Prediction Using Harris Hawks Optimization and Feature Analysis...

Area of Science:

• Computational biology
• Medical informatics
• Epidemiology

Background:

• The COVID-19 pandemic highlighted the need for advanced healthcare monitoring systems.
• Artificial intelligence (AI) and machine learning (ML) are crucial for analyzing big data generated during pandemics.
• Early detection of COVID-19 is vital for effective disease control and patient management.

Purpose of the Study:

• To propose an improved ML framework for the early detection of COVID-19.
• To optimize ML hyperparameters using the Harris Hawks Optimization (HHO) algorithm.
• To enhance prediction performance through an ensemble technique.

Main Methods:

• Applied HHO algorithm to optimize hyperparameters of ML models: eXtreme gradient boosting (XGBoost), light gradient boosting, categorical boosting, random forest, and support vector classifier.
• Utilized an ensemble technique combining optimized ML models for improved prediction.
• Evaluated feature importance using SHapely adaptive exPlanations (SHAP) values.

Main Results:

• The proposed ensemble ML model achieved a prediction accuracy of 92.38% on publicly available COVID-19 big data.
• The HHO-optimized eXtreme gradient boosting (HHOXGB) model showed the highest single-model accuracy at 92.23%.
• The proposed method demonstrated superior performance compared to traditional and other ML-based approaches.

Conclusions:

• The developed ML framework effectively improves early COVID-19 detection accuracy.
• Feature importance analysis provides insights into key indicators for disease detection.
• A graphical user interface is proposed for accessibility by non-specialist healthcare professionals.