Machine learning-based COVID-19 diagnosis by demographic characteristics and clinical data

Area of Science:

• Computational biology
• Medical informatics
• Epidemiology

Background:

• Effective COVID-19 screening is crucial for healthcare systems.
• Machine learning offers potential for rapid and accurate diagnosis.
• A user-friendly interface can aid healthcare professionals.

Purpose of the Study:

• To develop a machine learning-based prediction model for COVID-19 diagnosis.
• To design a graphical user interface (GUI) for symptom and demographic data input.
• To facilitate early clinical decision-making during the COVID-19 outbreak.

Main Methods:

• Implemented classification models: Support Vector Machine (SVM), Decision Tree (DT), Naïve Bayes (NB), K-Nearest Neighbor (KNN).
• Trained models on 16,973 individuals and tested on 1,885.
• Utilized Maximum Relevance Minimum Redundancy (MRMR) for feature selection and developed a GUI for prediction.

Main Results:

• Coughing showed the highest positive correlation with positive COVID-19 test results.
• The SVM model demonstrated the best performance with an accuracy of 70.48%, AUROC of 0.7045, and F1-score of 0.7157.
• The GUI integrates 42 features and symptoms for prediction using selected classification models.

Conclusions:

• A machine learning approach can facilitate early clinical decision-making for COVID-19.
• The predictive model effectively categorizes individuals into infected and non-infected groups.
• The developed system serves as an efficient screening tool for COVID-19 diagnosis.