Classification of the Human Development Index in Indonesia Using the Bootstrap Aggregating Method

Abstract

Successful development of the quality of human life in a region is determined by the Human Development Index (HDI). Human development performance based on the HDI can be measured: long and healthy life, knowledge, and a decent standard of living. The HDI is usually grouped into several categories to facilitate the classification of the HDI level of each region. This study aimed to determine the ability of the bootstrap aggregating (bagging) method to classify the HDI by district/city. Bagging is a stochastic machine learning approach that can eliminate the variance of the classifier by producing a bootstrap ensemble to obtain better accuracy results. The dependent variable in this study was the HDI by district/city in 2020. In contrast, life expectancy at birth, expected years of schooling, mean years of schooling, and real expenditure per capita are adjusted as independent variables. Bagging was applied to the high and low categories of HDI data. The bagging method demonstrated good classification performance due to only eight classification errors, namely the HDI data which should be in the high category but classified into the low category by the bagging method. Based on the results of calculations with 25 replications, it can be concluded that the bagging method has a very good performance, with an accuracy value of 92.3%, the sensitivity of 100%, and specificity of 83.33%. The bagging method is considered very good for the classifying the HDI by district/city in Indonesia in 2020 because it has a balanced accuracy of 91.67%.