Robust Principal Component Trimmed Clustering of Indonesian Provinces Based on Human Development Index Indicators

Abstract

Cluster analysis is a multivariate technique for grouping observations into clusters based on the observed values of several variables for each individual. The existence of outliers in the data can heavily influence standard clustering methods, i.e. the outliers will cause the standard clustering results to be not optimal. Therefore, it is necessary to use a robust clustering method. Trimmed clustering is one of robust clustering methods which is non-hierarchical and known for its good performance in cluster analysis when data contain outlier. The purpose of this study is to classify 34 provinces in Indonesia based on the 2019 Human Development Index (HDI) indicators and see the achievements of human development in each province. The results of this study indicate that there are three optimal clusters. The first cluster consists of 17 provinces with good HDI criteria, the second cluster consists of 9 provinces with a fairly good HDI, and the third cluster consists of 7 provinces with the lowest HDI criteria.