Analysis of COVID-۱۹ Distribution in Countries Using Unsupervised Machine Learning

Background and aims: Although more than two years have passed since the spearing of COVID-۱۹, a global pandemic, there are still major peaks in the number of confirmed cases and deaths.Although governments are trying to overcome this disease with different policies, it is not entirelycontrolled yet. Therefore, many researches have been conducted to address this issue. However,most of them considered individuals’ features such as disease symptoms. The current study aimsto analyze data at the country level and use health, economic, and social information in order tocluster countries affected by COVID-۱۹ employing machine learning. Thus, countries with similarfactors can take proactive steps to control the pandemic.Method: In this study, two methods of unsupervised learning, K-Means and Hierarchical algorithms,were used to cluster ۲۰۷ countries based on factors such as the number of confirmedCOVID-۱۹ cases, deaths, vaccinated people, handwashing facilities, and other socio-economicindicators like GDP per capita, smoking prevalence, and life expectancy. The optimal number ofclusters was considered k=۶ based on the elbow method. To obtain the most associated features,the correlation between selected variables and confirmed COVID-۱۹ cases, deaths, and vaccinationrates were analyzed.Results: The results revealed that countries in the same clusters have behaved similarly in dealingwith this pandemic. For example, Russia, Argentina, South Korea, and Italy are in the samecluster with a relatively low stringency index and as a result, a low number of vaccinations. Countriessuch as Canada, Sweden, and Egypt where the human development index, life expectancy,and the amount of GDP per capita were the highest placed in the same cluster. In this cluster,relatively more people were vaccinated and the mortality was low.Conclusion: The government stringency index showed a strong correlation with the number ofvaccinations, whereas environmental health indicators were weakly correlated with mortalityfrom COVID-۱۹. Politicians can make better decisions by considering these indicators and therefore,manage the negative consequences of COVID-۱۹.