Abstract
COVID-19, a pandemic, attacked millions of people’s health and economies across the world, particularly in low-income developing countries such as Pakistan. The study aims to develop a novel method and approach to diagnose COVID-19. Clinical features C-reactive protein, ferritin, and D-dimer levels were accessed to check the severity of COVID-19 positive patients. 160 patients were included in this study who had positive signs for COVID-19. Sandwich immune-detection and real time-PCR analyses were performed to access the clinical features of COVID-19. The results of clinical features and real time-PCR assay were compared using Artificial Intelligence (AI). Four classifiers; Support vector machine, Random Forest, K- nearest neighbor, and Neural network, were used to predict the results and the accuracy from these algorithms was 78.6%, 75.4%, 75.4%, and 63.9% respectively. The higher accuracy was from the Support vector Machine which shows 78.6% accuracy of clinical features results obtained from COVID-19 positive patients. In conclusion, this study provides an alternative diagnostic method for COVID-19 patients. Additionally, this study not only provided the diagnostic method but also evaluate severity of clinical features and also the cost-effective diagnosis of COVID-19 detection. The alternative way provided by this this study will be very helpful for the diagnosis of COVID-19 through basic test parameters.