Machine Learning Algorithms for Predicting Stunting among Under-Five Children in Papua New Guinea

Area of Science:

• Public Health
• Pediatrics
• Machine Learning in Health

Background:

• Childhood stunting remains a persistent challenge in Papua New Guinea (PNG), impacting long-term health and development.
• Effective prediction models are needed to identify at-risk children for targeted interventions.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting stunting in children under five in PNG.
• To identify the most significant predictors of stunting using advanced analytical techniques.

Main Methods:

• Utilized data from the 2016-2018 PNG Demographic Health Survey (n=3380).
• Employed feature selection (LASSO, RF-RFE) and predictive modeling (Logistic Regression, Conditional Decision Tree, SVM, XGBoost).
• Evaluated model performance using accuracy, precision, recall, F1 score, and AUC, with SHAP values for predictor importance.

Main Results:

• The LASSO-XGBoost model demonstrated superior performance (AUC: 0.765) for stunting prediction.
• Key predictors identified include living in the Highlands Region, child's age, richest family wealth quintile, and birth size.
• The model achieved an accuracy of 0.728 and an F1 score of 0.669.

Conclusions:

• Machine learning offers a powerful tool for predicting childhood stunting in PNG.
• Early identification of high-risk factors can guide targeted nutritional and health interventions.
• Focusing on maternal and child nutritional status is essential for preventing stunting and improving overall well-being.