Using Machine Learning Approaches to Distinguish COVID-19 from Other Upper Respiratory Tract Infections

ABSTRACT

COVID-19 has made its first debut in early December 2019 in Wuhan, China. The COVID-19 main symptoms are fever, sore throat, tiredness, and cough, which are quite similar to flu, cold, and allergic rhinitis symptoms. COVID, flu, allergic rhinitis, and cold are all caused by respiratory attacking viruses, and they transmit in the same way by droplets and contacting surfaces and bodies. Thereby, it becomes an urgent need for physicians and healthcare providers to differentiate between each case of the four diseases of cold, flu, allergic rhinitis, and COVID-19. Making an accurate diagnosis in a timely manner, maintain that the patient has the best chance to get to the best clinical outcome. Thus, to meet this aim, the current paper used a publicly available symptoms dataset and applied the Apriori algorithms to extract the most important rules between the symptoms. Furthermore, six classifying algorithms were used to predict the type of disease from its' symptoms, these algorithms namely;Bagging, Random Forest, Extra Trees, Ada Boost, Stochastic gradient boosting, and voting ensemble. The experiments show that the voting ensemble algorithm achieved the highest classification testing accuracy (about 96.22%). As a conclusion, we conclude by demonstrating that using an ensemble technique may greatly increase classification accuracy and make COVID-19 easier to distinguish from other similar diseases. For the medical sector and the healthcare business, our results have significant theoretical and practical consequences. © 2022 IEEE.