Application of machine learning algorithm for predicting acute malnutrition among under 5 children in east Africa using recent DHS.

Acute malnutrition remains a critical public health challenge across East Africa, contributing substantially to under-five morbidity and mortality. Early identification of at-risk children using predictive models could enhance timely intervention. This study aimed to develop and evaluate machine learning models for predicting acute malnutrition among under-five children in East Africa. A cross-sectional analysis was conducted using the most recent DHS data from 12 East African countries. A total of 76,224 children under-five years were included. Data preprocessing involved multiple imputation for missing values, one-hot encoding for categorical variables, SMOTE to address class imbalance, and multiple feature selection. ten supervised machine learning algorithms; Logistic Regression, Random Forest, K-Nearest Neighbors, Decision Tree, Support Vector Machine, Naïve Bayes, XGBoost, LightGBM, CatBoost, and Stochastic Gradient Descent were trained and evaluated using an 80/20 train-test split and stratified five-fold cross-validation. Model performance was assessed using accuracy, AUC, precision, recall, and F1-score, and interpretability was explored using SHapley Additive exPlanations (SHAP). The overall prevalence of acute malnutrition across east Africa was 6.08% (95% CI 5.91-6.25%), ranging from 1.2% in Rwanda to 12.27% in Ethiopia. The Random Forest algorithm demonstrated the best predictive performance, achieving an AUC of 74.6% (95% CI 73.0-76.2), accuracy of 71.2%, precision of 13.1%, recall of 66.2%, specificity of 71.5%, and an F1-score of 0.218. SHAP analysis identified important predictors of acute malnutrition as rural residence, large family size (≥ 6 members), higher number of under-five children (≥ 3), small or very small birth size, inadequate antenatal care, unimproved water sources, inappropriate disposal of child feces, recent diarrheal illness, and lack of latrine facilities. Protective factors included maternal literacy, antenatal care attendance, vitamin A supplementation, longer birth intervals (≥ 2 years), and maternal employment. Random Forest, demonstrate superior predictive capacity for identifying children at risk of acute malnutrition in East Africa. The integration of such models into national nutrition surveillance systems could enable early detection and targeted interventions. Strategies should prioritize rural communities, strengthen maternal education and empowerment, improve water and sanitation infrastructure, and antenatal care to mitigate risk factors and enhance child nutritional outcomes.