Clustering of regional HDI data using Self-Organizing Maps

Abstract

The Municipal Human Development Index (HDI) was established by the United Nations Development Program. It uses data from longevity, education, and income to infer regional social and economic life quality. This work developed a multivariate analysis of the HDI data evolution in years 1991, 2000 and 2010 of 167 municipalities in Rio Grande do Norte State, northeast Brazil. Self-organizing (Kohonen or SOM) maps were used to perform clustering and data visualization. The approach uses map segmentation with k-means algorithm after SOM training. Transition analysis from municipalities in different studied years are performed, presenting ranking of clusters in terms of the three main HDI dimensions. Five groups of municipalities resulted from intragroup similarities and intergroup dissimilarities in each period. The segmented maps present similar municipalities. Thematic maps of the region after data clustering are also shown.