Clustering and Correlation Methods for Predicting Coronavirus COVID-19 Risk Analysis in Pandemic Countries

ABSTRACT

An extraordinary outbreak of pneumonia in Wuhan City, China, was subsequently termed as COVID-19 emerged in December 2019. The virus is also known as an infectious disease inherited from a novel coronavirus. This study exposed the beginning of the unprecedented COVID-19 confirmed cases spike exponentially in the United States and 200 countries globally. Epidemiologists usually utilize conventional spread prediction via the classic clustering method. A suspected patient is likely to blow out the disease to a potential agglomerative of cases grouped in place and time. In the era of cutting edge, outbreak prediction can also generate accurate techniques to utilize unsupervised machine learning methods. We apply two prominent unsupervised learning methods, namely K-means clustering and correlation on a set Coronavirus Outbreak COVID-19 data collection dated March 27 and August 16, 2020. The K-means automatically search for unknown clusters of many countries infected with the COVID-19 rapidly. It shows that a group of $m = 5$ produces an accuracy of about 97% with [The United States and Italy], [Iran, France], [Spain, German], [Indonesia, Malaysia, Philippine] as clusters. At the same time, it predicts a pertinent relationship between the total deaths and critical patients' attributes of 0.85 while correlating COVID-19 characteristics. © 2020 IEEE.