Answers are grounded in this paper and cite their sources.
Background Severe child food poverty is a significant nutritional concern, particularly during the complementary feeding period of 6–23 months, when inadequate dietary diversity can adversely affect child growth and development. Identifying children at greater risk may support timely and targeted nutrition interventions. This study aimed to predict severe child food poverty using machine learning models and identify its associated factors among children aged 6–23 months in Ethiopia. Methods A nationally representative sample of 3,792 children aged 6–23 months from the 2024-25 Ethiopia Demographic and Health Survey was analyzed. Severe child food poverty was defined as consumption of foods from 0–2 of eight recommended food groups during the preceding 24 hours. Descriptive and inferential analyses were conducted to identify associated factors. Seven machine learning algorithms—Random Forest, XGBoost, Decision Tree, Artificial Neural Network, LightGBM, K-Nearest Neighbors, and Logistic Regression—were developed and evaluated. The training data were balanced using SMOTE, while tenfold cross-validation and RandomizedSearchCV were used for model tuning. Model performance was assessed using accuracy, precision, recall, F1-score, ROC-AUC, and PR-AUC. Results Among the 3,792 children aged 6–23 months, 1,780 (46.94%) experienced severe child food poverty. Child age, mass-media exposure, maternal education, wealth index, maternal age household size, and antenatal care attendance were significantly associated with severe child food poverty. Children aged 6–11 months, those from poorer households and larger households, and those born to younger or less-educated mothers had higher likelihood of severe food poverty. In contrast, greater maternal mass-media exposure and adequate ANC attendance were associated with lower likelihood of severe food poverty. Among the evaluated machine learning models, Random Forest showed the best predictive performance, achieving 93.54% accuracy, 93.61% precision, 93.49% recall, and 93.53% F1-score, followed by XGBoost and Decision Tree. Conclusion Severe child food poverty affected nearly half of children aged 6–23 months in Ethiopia. The strong performance of Random Forest demonstrates the potential of machine learning to identify children at high risk of severe food poverty and to capture complex relationships among maternal, household, and child-related factors. Integrating predictive approaches with targeted nutrition interventions may help improve dietary adequacy during the critical complementary feeding period.
No takes yet. Share an insight, caveat, or question.
Tewoligne et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: