Accurate Classification of COVID-19 Based on Incomplete Heterogeneous Data using a KNN Variant Algorithm

Abstract The coronavirus 2019 disease (COVID-19) is wreaking havoc around the world, and great efforts are underway to control it. Millions of people are now being tested and their data keeps accumulating in large volumes. This data can be used to classify newly tested persons as whether they have the disease or not. However, normal classification techniques are hampered by the fact that the data is typically both incomplete and heterogeneous. To address this two-fold obstacle, we propose a KNN variant (KNNV) algorithm which accurately and efficiently classifies COVID-19. The main two ideas behind the proposed algorithm are that for each instance to be classified it chooses the parameter K adaptively and calculates the distances to other instances in a novel way. The KNNV was implemented and tested on a COVID-19 dataset from the Italian society of medical and intervention radiology society. It was also compared to three algorithms of its category. The test results show that the KNNV can efficiently and accurately classify COVID-19 patients. The comparison results show that the algorithm greatly outperforms all its competitors in terms of four metrics: precision, recall, accuracy, and F-Score.