Machine Learning for Child Malnutrition Prediction in Indonesia

This study utilizes machine learning (ML) methods to predict malnutrition among children in Indonesia. It adopts a family-centered perspective, leveraging large datasets from WHO, UNICEF, and the Global Nutrition Report. A 100,000-strong stratified sample of children aged 0-18 was examined, incorporating a set of anthropometric, socioeconomic, and demographic factors. Preprocessing of data included imputation, normalization, and feature selection. Results revealed a dual burden of malnutrition, with stunting proportion being 25.3% and overweightness proportion being 8.6%, and regional disparities indicating higher proportions in rural areas and provinces such as Aceh and South Kalimantan. Analysis of feature importance identified weight-for-age, parental education, household income, and access to clean water as key predictors of health outcomes. The model had peak performance for children aged 6-10 years. These findings highlight the strength of ML to improve health surveillance, inform targeted nutritional interventions, and enhance evidence-based policymaking. The framework provides actionable insights for enhancing national initiatives, such as Bangga Kencana, so that family planning efforts are aligned with overall child health targets. Future studies should focus on improving data quality for rural environments, adding environmental and dietary factors into models, and exploring advanced ensemble models for higher generalizability and applicability to policy.