Machine learning algorithms for predicting undernutrition among under-five children in Ethiopia.

Area of Science:

• Public Health
• Machine Learning Applications
• Global Health

Background:

• Child undernutrition, including stunting, wasting, and underweight, is a critical global public health issue.
• Ethiopia faces significant regional variations in child undernutrition rates.
• Identifying determinants of undernutrition is crucial for effective public health interventions.

Purpose of the Study:

• To estimate predictive algorithms for childhood stunting determinants using machine learning (ML).
• To identify socio-demographic risk factors associated with undernutrition in Ethiopian children.
• To compare the predictive performance of various ML algorithms for child undernutrition.

Main Methods:

• Utilized data from the 2016 Ethiopian Demographic and Health Survey.
• Applied five ML algorithms: eXtreme gradient boosting (xgbTree), k-nearest neighbors (k-NN), random forest, neural network, and generalized linear models.
• Analyzed data from 9471 children under five years of age.

Main Results:

• The xgbTree algorithm demonstrated superior predictive ability compared to other ML models.
• Key predictors of undernutrition included time to water source, anemia history, child age, birth size, and maternal underweight.
• Substantial regional disparities in stunting, wasting, and underweight were observed.

Conclusions:

• The xgbTree algorithm is a highly effective ML tool for predicting childhood undernutrition in Ethiopia.
• Findings highlight the importance of improving water supply, food security, and fertility regulation for child nutrition.
• Interventions targeting identified risk factors can significantly improve childhood nutrition outcomes in Ethiopia.