Modeling Human Development Index of East Java Using Spatial Autoregressive and Spatial Error Ensemble

Abstract

The human development index (HDI) is an indicator used to monitor the government's success in developing the quality of human life. East Java Province's HDI is the lowest compared to other provinces on Java Island. Therefore, it is necessary to improve human development in this province. Attention must be paid to all aspects of human development, including the relationship between neighboring regions. The spatial regression method is an analysis method that considers the spatial dependency of the data. Ensemble spatial regression combines several spatial models by adding noise to the response variable, which is expected to reduce the diversity in the data. This research aims to use ensemble spatial regression to examine the East Java HDI. East Java HDI has spatial lag and spatial error dependence, modeled with SAR and SEM. Queen contiguity is used as a spatial weight. The SEM model does not fulfill the homogeneity assumption, so it is continued with the ensemble method. The ensemble method is proven to reduce diversity, so  SEM Ensemble fulfills the assumption of homoscedasticity. After analysis using SAR and SEM Ensemble, the SAR model was chosen as the best model with the largest  and lowest AIC value. Significant variables on East Java HDI are life expectancy, expected years of schooling, average years of schooling, and expenditure per capita.