Detection COVID-19 using Machine Learning from Blood Tests

Coronavirus (COVID-19) is an infectious disease that spreads around the world lately. It is a worldwide pandemic that affects mainly the lungs. It leads to huge risks as blood clotting, heart attack, and most of the time leading to death if it is not detected and cured early. The purpose of this research is to detect COVID-19 in its early stages to minimize the number of deaths caused by this disease. Discovering COVID-19 in early stages minimize the economy loss according to its drastic effect in manpower production. For the first time, detecting COVID19 occurrence using blood tests which are easy to collect and at the same time not expensive. We generated our own dataset which was used in detecting COVID-19. The dataset includes 134 cases from different genders, different ages and separated between positive and negative cases. The dataset contains blood tests which are (CBC, CRP, D-Dimer, S-ferritin, ALT,LDH). By these blood tests, doctors can know if people are infected by COVID-19 or not. Our approach used Machine Learning technique with different classifiers such as Random Forest (RF), Support Vector Machine (SVM) and Naive Bayes. The accuracy achieved were 76% for RF, 88% for SVM and 85% for Naive Bayes. SVM classifier achieved the best accuracy and was used in our model. According to the American Center of Disease Control and prevention (CDC), the accuracy of PCR is 80%, and it is expense relative to the blood tests, and using our model we achieved a better results than the PCR.