Abstract

In this study, principal component analysis and the k-means algorithm were employed for the analysis of Mortality-Based Global Health Metrics. The aim of this study is to create homogeneous clusters of countries in terms of MortalityBased Global Health Metrics, to identify similar countries within clusters using within-cluster exploratory data analysis methods, and to investigate the common characteristics of these countries. At the country level, a dataset comprising 34 indicators was compiled. However, due to the curse of dimensionality inherent in machine learning, the dataset was reduced to 6 principal components through principal component analysis (PCA). Countries were then clustered into 6 groups using the K-means clustering analysis method. The elbow method and silhouette method were utilized for optimal cluster selection. The cluster information resulting from dimensionality reduction analysis and clustering analysis can serve as a valuable input for policymakers in healthcare, particularly regarding cluster centroids and the countries constituting each cluster. Healthcare policymakers for each country can develop much more rational policies in their decisionmaking processes by evaluating their own countries, other countries within the same cluster, the characteristic features of their own clusters, and the distances to successful cluster centroids. This enables better examination of positive and negative indicators in country comparisons.

Keywords: Health, Mortality, World Development Indicators, Principal Component Analysis, Clustering