Predicting stunting status among under five children in ethiopia using ensemblemachine learning algorithms.

Area of Science:

• Public Health
• Machine Learning
• Pediatrics

Background:

• Childhood stunting remains a critical public health issue in Ethiopia, affecting child development and well-being.
• Previous prediction models had limitations in addressing stunting severity.
• Ethiopian Demographic and Health Surveys (EDHS) provide valuable national data for analysis.

Purpose of the Study:

• To develop a multi-class classification model for predicting childhood stunting severity (severe, moderate, normal) in Ethiopia.
• To identify key risk factors associated with childhood stunting.
• To leverage machine learning for improved stunting prediction and intervention planning.

Main Methods:

• Analysis of secondary data from the 2011 and 2016 Ethiopian Demographic and Health Surveys (EDHS).
• Implementation of data preprocessing techniques including SMOTE for class balancing.
• Evaluation of four ensemble machine learning algorithms: Random Forest, AdaBoost, XGBoost, and CatBoost.

Main Results:

• The Random Forest model demonstrated superior performance with high accuracy (97.985%) and ROC-AUC (99.995%).
• Identified significant risk factors include child's age, maternal education, birth order, wealth index, maternal BMI, breastfeeding duration, and access to clean water and sanitation.
• The model effectively predicted stunting severity across multiple classes.

Conclusions:

• Machine learning, particularly Random Forest, is highly effective for predicting childhood stunting in Ethiopia.
• Findings offer crucial data for developing targeted interventions to combat childhood stunting.
• The study highlights the importance of addressing socioeconomic and health-related factors to reduce stunting prevalence.