THE USE OF EXPLAINABLE AI FOR ANALYZING SOCIOECONOMIC DETERMINANTS OF THE HUMAN DEVELOPMENT INDEX IN INDONESIA BASED ON REGRESSION MODELS

Abstract

The Human Development Index (HDI) is a key indicator of quality of life, reflecting achievements in health, education, and a decent standard of living. Significant regional disparities in Indonesia highlight the need to analyze its determinants for effective policy formulation. This study examines the simultaneous influence of socioeconomic factors—poverty rate, GRDP per capita, life expectancy, mean years of schooling, and expenditure per capita—on HDI across 514 regencies/cities using machine learning and Explainable AI (XAI). Secondary data from the Indonesian Central Bureau of Statistics (BPS) in 2021 were utilized. The target variable (IPM_score) was constructed through feature engineering. Linear Regression, Random Forest, and XGBoost models were trained using an 80:20 split and evaluated with Mean Squared Error (MSE) and R². SHAP was applied to interpret feature contributions. Results show XGBoost achieved the best performance (R² = 0.987), outperforming Random Forest (R² = 0.974), while Linear Regression achieved R² = 1.000 due to perfect linearity. SHAP analysis identified expenditure per capita as the most dominant factor (r = 0.9996), followed by mean years of schooling (r = 0.667), while poverty showed a strong negative effect (r = -0.638). These findings emphasize that purchasing power and education are critical drivers of HDI. The use of XAI enhances model transparency and supports evidence-based policy, particularly in integrating poverty reduction with improvements in education and economic capacity.