Exploring Risk Factors and Predictive Modeling of Child Malnutrition in Pakistan Using Machine Learning

Area of Science:

• Pediatric Nutrition
• Computational Epidemiology
• Public Health Informatics

Background:

• Child malnutrition, encompassing stunting, wasting, and underweight, remains a critical public health issue in Pakistan.
• Socio-demographic and health-related factors significantly influence malnutrition prevalence in children under five.

Purpose of the Study:

• To identify risk factors associated with child malnutrition in Pakistan.
• To develop and evaluate machine learning models for predicting child malnutrition (stunting, wasting, underweight).
• To inform targeted public health interventions.

Main Methods:

• Cross-sectional analysis of Pakistan Demographic and Health Survey (PDHS) 2017-2018 data.
• Logistic regression for risk factor identification.
• Application and performance evaluation of Random Forest, SVM, Naïve Bayes, and AdaBoost models using cross-validation and train-test splits.

Main Results:

• Consanguineous marriages, lower wealth, low maternal education, and regional disparities (Sindh, Baluchistan) were identified as significant risk factors.
• Random Forest model achieved the highest accuracy and specificity for predicting malnutrition indicators.
• Support Vector Machine (SVM) demonstrated superior sensitivity for wasted and underweight children.

Conclusions:

• Child malnutrition in Pakistan is linked to a complex interplay of socio-demographic factors.
• Machine learning offers a powerful approach for predicting child malnutrition.
• Interventions should prioritize maternal education, sanitation, and poverty reduction in high-risk areas.