| Abstrak/Abstract |
This study aims to compare the quality of nine clustering methods in classifying 180 countries. The rates compared are the homogeneity of groups, and the significance of differences of vector means. We apply hierarchical clustering with linkage of average, centroid, complete, median, McQuitty, and single, and Ward’s method. We also implement non-hierarchical partition methods: k-means and the block-based k-medoids. This research used secondary data about the 2022 Environmental Performance Index (EPI). This study uses three variables of EPI: Environmental Health, Ecosystem Vitality and Climate Change. A medoid-based deviation ratio index determines the number of clusters, which concludes three groups as the optimal number of classes. The hierarchical grouping method that produces the best homogeneity quality is the Ward method. Meanwhile, the best homogeneity quality produced by a non-hierarchical approach is the block-based k-medoids partition method. The block-based k-medoids partitioning method produces mean vector comparisons between groups with a more understandable interpretation than Ward’s method. The best group on the indicators for Environmental Health, Environmental Vitality and Climate Change are the countries mainly from the Global West and Eastern Europe regions. |
