Supervised Machine Learning Approach to COVID-19 Detection Based on Clinical Data

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Epidemiology

Background:

• The COVID-19 pandemic necessitated rapid, non-clinical solutions for diagnosis and control.
• Artificial intelligence (AI) techniques offer potential for early detection and reducing healthcare system burden.

Purpose of the Study:

• To evaluate the efficacy of Machine Learning (ML) algorithms for the early detection of COVID-19.
• To compare the performance of six different ML algorithms in diagnosing COVID-19.

Main Methods:

• Retrospective study utilizing a dataset of 10055 COVID-19 cases with 63 features (March-October 2020).
• Comparison of C4.5, Support Vector Machine (SVM), Naive Bayes, Logistic Regression (LR), Random Forest, and K-Nearest Neighbor algorithms using Rapid Miner.
• Performance evaluation based on precision, recall, accuracy, and f-measure.

Main Results:

• Support Vector Machine (SVM) achieved 93.41% accuracy, and C4.5 achieved 91.87% accuracy.
• The C4.5 decision tree identified "contact with a person who has COVID-19" as the most significant diagnostic criterion (Gini index).

Conclusions:

• Machine learning approaches demonstrate a reasonable level of accuracy for COVID-19 diagnosis.
• ML algorithms can serve as valuable tools for early COVID-19 detection and epidemic management.