Multidimensional Poverty Clustering using K-Means Algorithm with Dimensionaly Reduction by Principal Component Analysis

Abstract

The level of Multidimensional poverty in each province in Indonesia varies, similar policies is ineffective to reduce the poverty. Several poverty indicators also influence other factors. General policies established to overcome poverty have proven ineffective, making it urgent to identify the needs of each province in overcoming this condition. Grouping provinces based on similar multidimensional poverty which use cluster analysis, will help address this situation. The aim of this study is to group provinces based on multidimensional poverty indicators using the k-means clustering method. Principal Component Analysis (PCA) was also used to reduce variables and multicollinearity. The clustering results showed seven clusters. The highest multidimensional poverty was found in cluster 2, which consisted of one province, namely Papua Pegunungan. This province shows deficiencies in education, health, and living standards compared to other clusters. Meanwhile, the lowest multidimensional poverty was found in cluster 7. There are three provinces in this cluster, namely Bali, Jakarta, and DIY Jogjakarta. These provinces experience minimal multidimensional poverty which is able to provide a better quality of life. The policies and development strategies in these provinces could serve as role models to develop other provinces based on their specific deficiencies and needs.   Each cluster is well separated, as Davies Bouldin Index (DB) is lover, at 0.4.