IDENTIFICATION OF STUNTING CAUSES USING MACHINE LEARNING ON HEALTH AND SOCIO-ECONOMIC DATA IN INDONESIA

Abstract

Stunting is a multifaceted nutritional issue that is impacted by the intricate interplay between family structural constraints and child health concerns. This study uses a machine learning approach to simulate the prevalence of stunting and determine the significant ranking of its major causes in Indonesia. Ten independent variables from the socioeconomic and health dimensions are integrated using the Random Forest Regressor technique. With an R-Squared value of 0.89, the model evaluation results demonstrated a very high level of performance, demonstrating the algorithm's dependability in mapping data variance. With a contribution of 36.36%, the Percentage of Poor Population was shown to be the most significant predictor based on the extraction of the Feature Importance measure, followed by Complementary Feeding (18.36%) and Per Capita Expenditure (12.75%). When taken as a whole, these results demonstrate that the incidence of stunting is significantly more influenced by economic welfare determinants than by clinical health interventions alone. According to the study's findings, stunting is mostly caused by poverty and low purchasing power, necessitating comprehensive policy approaches centered on reducing cross-sectoral structural poverty rather than merely focusing on particular medical treatments.