Machine learning techniques as an efficient alternative diagnostic tool for COVID-19 cases

Background: The SARS-CoV-2 virus has demonstrated the weakness of many health systems worldwide, creating a saturation and lack access to treatments. A bottleneck fight this pandemic relates diagnostic infrastructure for early detection positive cases, particularly in rural impoverished areas developing countries. In context, less costly fast machine learning (ML) diagnosis-based are helpful. However, most research focused on deep-learning techniques diagnosis, which computationally technologically expensive. ML models have been mainly used as benchmark not entirely explored existing literature topic paper. Objective: To analyze capabilities (compared deep learning) diagnose COVID-19 cases based X-ray images, assessing performance these using their predictive power such diagnosis. Methods: factorial experiment was designed establish with chest images healthy, pneumonia, infected patients. This design considers data-balancing methods, feature extraction approaches, different algorithms, hyper-parameter optimization. were evaluated classification metrics, including accuracy, area under receiver operating characteristic curve (AUROC), F1-score, sensitivity, specificity. Results: provided mean its confidence intervals capability techniques, reached AUROC values high 90% suitable sensitivity Among support vector machines random forest performed best. down-sampling method unbalanced data improved significantly study. Conclusions: Our investigation that able identify results specificity, minimizing false-positive or false-negative rates. trained low computational resources, helps provide deployment areas.