Clustering Countries of the World Based on the Trend of the COVID-19 Incidence: An application of shape-based k-means algorithm

Abstract

Background: the urban Health Commission of Wuhan City, China, issued an emergency notice due to an incidence of viral pneumonia of unknown cause in December 2019. The World Health Organization officially named it the 2019 novel coronavirus. Since the course of the disease is not the same in the countries and regions of the world, and the study of this diversity is an important source of information for policymakers and researchers. The current study aims to cluster selected countries of the world based on its incidence and it was done according to the trajectories shape. Methods: the data set analyzed, included new cases of COVID-19(per million people) in 13 countries of the world, which were published on the World Health website from March, 2020, to October, 2021 monthly then analyzed by the k-means clustering method using Fréchet distance and R software V4.0.5. in addition, clustercrit and kmlshape packages were utilized for trajectory clustering. Result: Research results show that, 13 countries of the world were classified into 2 clusters with high and low incidence. The cluster with high incidence included 8(62%) countries. The 2 cluster exhibits the highest outlier in COVID-19 incidence. In the analyzed nations, United States and Brazil exhibited the highest incidence rate in clusters 1 and 2, respectively. Conclusion: The present findings showed that there are 2 patterns in the epidemic of COVID-19. The first pattern includes severe fluctuations and the next pattern includes low fluctuations. The results revealed that the method used in this article has the potential to understand incidence trends regardless of the time of disease onset. Since the Covid-19 infection process experiences fluctuations over time that vary based on when the pandemic began in each country, it's essential to analyze the similarity and shape of infection trends collectively, independent of their starting points.

Keywords

• Clustering, COVID-19, K-means, Fréchet distance, Longitudinal data, incidence rate