Implementation of the trimmed k-means clustering method in mapping the distribution of Covid-19 in Indonesia

Abstract

The Coronavirus that appeared in (COVID-19), caused by SARSCoV-2, started at Wuhan in the Hubei province of China and has spread with great speed around the world; it has caused a severe health crisis all around the world, including Indonesia. This study aims to use a clustering technique to assess the risk of the COVID-19 pandemic in Indonesia, based on data obtained between March 2020 and July 2021 in that country (http://www.covid19.go.id ). Provinces in Indonesia were grouped based on COVID-19 infection rates and mortality data. Since the data con-tained some outliers, i.e. provinces with a very high number of cases, we used a robust clustering method; this method is sensitive to outliers. The analysis was performed using the Trimmed k-means clustering method. Based on the results of this study, with four provinces detected as outliers in the data, there were three optimal clusters with the maximum separation index. Cluster 1 consisted of 14 provinces, and clusters 2 and 3 consisted of 10 and 6 provinces, respectively. The four outliers, i.e. Jakarta, West Java, Central Java and East Java, formed a separate cluster.