Machine learning-based COVID-19 diagnosis for an IoT centered community health center (Health kiosk)

Abstract

Since the outbreak of the novel coronavirus, many countries have set up testing centres across states to test and isolate people with the virus to reduce its spread. Others have taken the vaccination program seriously, with Israel vaccinating over 85% of its population to make them immune to this virus as of October 2021. Unfortunately, Ghana is behind in these two initiatives to curb the virus. The lack of testing centres in rural areas, coupled with the cost of testing, makes it hard for people to get tested. In this paper, we trained a machine learning model called the Gradient Boosting Machine to serve as an alternative testing system for diagnosing COVID-19. We used haematological Rhinovirus.Enterovirus status from 5,644 patients as dependent variables, and their COVID- 19 results as the independent variable, to train the model. The top five dependent variables with the most significant relative influence on the model are Rhinovirus.Enterovirus, Leucocytes, Influenza.B, Platelets and Inf.A.H1N1.2009. The model exhibited a higher sensitivity of 97.9%, a specificity of 78% and an AUC of 92%. This diagnostic accuracy can at least complement the RT-PCR test or be compared to it.