Robust k-means clustering for modeling the spread of coronavirus disease in Indonesia

The Indonesian government has taken several policies to break the chain of transmission of the COVID-19 virus infection since the spread of this virus variant was first discovered in March 2020 in Indonesia. One of them is to require everyone to get vaccinated. This national vaccination program has been started since early January 2021 until now and expected to be one of the solutions to control the spread of coronavirus in Indonesia. This paper presents robust k-means to show the cluster of the spread of COVID-19 in Indonesia after the implementation of the vaccine policy for Indonesian citizens. The k-means algorithm is an unsupervised machine learning method. This method is well known in clustering. This is a partition clustering algorithm which is widely used in practice but in general, this method is sensitive to the presence of outliers. The variables considered in this study are the number of deaths, the number of recovered patients and the number of patients. Experimental comparisons are made with the basic k-means and robust k-means algorithm. It seems that for data on May, 2021, the cluster with k=3 in the robust case will provide closer risk levels of the provinces in Indonesia with their risk in reality, namely provinces with high, mid and low-level risk.