An Ensemble Feature Selection Approach-Based Machine Learning Classifiers for Prediction of COVID-19 Disease

Area of Science:

• Medical Informatics
• Computational Biology
• Data Mining

Background:

• Coronavirus disease 2019 (COVID-19) poses a significant global health and economic threat.
• Effective control strategies rely heavily on early and accurate detection to limit transmission.
• The development of a definitive cure for COVID-19 is still pending, emphasizing the need for robust diagnostic tools.

Purpose of the Study:

• To propose and evaluate a novel data mining system for the effective identification of COVID-19 infection.
• To assess the performance of various feature selection methods in enhancing machine learning classifier accuracy for COVID-19 detection.
• To identify optimal features that support COVID-19 datasets for improved diagnostic capabilities.

Main Methods:

• An ensemble feature selection approach was developed, integrating chi-square test, Recursive Feature Elimination (RFE), Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Random Forest.
• Machine learning classifiers including Decision Tree, Naïve Bayes, K-nearest neighbor (KNN), Multilayer Perceptron (MLP), and Support Vector Machine (SVM) were employed.
• Two distinct COVID-19 datasets were utilized to test the proposed system and extract the most supportive features.

Main Results:

• The study evaluated the effectiveness of different feature selection techniques in improving the classification accuracy of various machine learning models.
• The performance analysis focused on how ensemble feature selection impacts the diagnostic capabilities of classifiers like SVM, KNN, and others.
• Extracted features were identified as crucial for enhancing the predictive power of the machine learning models on the COVID-19 datasets.

Conclusions:

• The proposed data mining system, combining ensemble feature selection and machine learning, demonstrates potential for effective COVID-19 identification.
• Feature selection plays a vital role in optimizing machine learning models for accurate detection of infectious diseases like COVID-19.
• Further research can build upon these findings to develop more advanced and reliable diagnostic systems for respiratory illnesses.