COVID-19 Diagnosis at Early Stage Based on Smartwatches and Machine Learning Techniques

Early detection of COVID-19 positive people are now extremely needed and considered to be one the most effective ways how limit spreading infection. Commonly used screening methods reverse transcription polymerase chain reaction (RT-PCR) or antigen tests, which need periodically repeated. This paper proposes a methodology for detecting disease in non-invasive way using wearable devices analysis bio-markers artificial intelligence. have reused publicly available dataset containing COVID-19, influenza, Healthy control data. In total 27 healthy were pre-selected experiment, several feature extraction applied experimented with machine learning methods, such as XGBoost, k-nearest neighbour k-NN, support vector machine, logistic regression, decision tree, random forest, statistically evaluated their perfomance various metrics, including accuracy, sensitivity specificity. The proposed experiment reached 78 % accuracy k-NN algorithm is significantly higher than reported state-of-the-art methods. For cohort was 73 k-NN. Additionally, we identified relevant features that could indicate changes between infected state. can complement existing RT-PCR it help viral diseases, not only way.