Machine learning algorithms for predicting malnutrition among under-five children in Bangladesh

Area of Science:

• Pediatric Nutrition
• Machine Learning Applications
• Public Health Informatics

Background:

• Malnutrition remains a significant public health challenge for children under five in Bangladesh.
• Accurate and timely prediction of malnutrition status is crucial for effective intervention strategies.

Purpose of the Study:

• To evaluate and compare the efficacy of various machine learning (ML) algorithms in predicting malnutrition status among under-five children in Bangladesh.
• To identify the most effective ML model for this predictive task.

Main Methods:

• Utilized secondary data from the 2014 Bangladesh Demographic and Health Survey (BDHS).
• Applied and assessed five ML algorithms: Linear Discriminant Analysis (LDA), k-Nearest Neighbors (k-NN), Support Vector Machines (SVM), Random Forest (RF), and Logistic Regression (LR).
• Evaluated algorithm performance using accuracy, sensitivity, specificity, and Cohen's kappa statistic.

Main Results:

• The Random Forest (RF) algorithm achieved the highest accuracy (68.51%), sensitivity (94.66%), and specificity (69.76%).
• RF classification demonstrated the strongest discriminative ability with a Cohen's kappa of 0.2434.
• RF outperformed other evaluated ML algorithms in predicting malnutrition status.

Conclusions:

• The Random Forest algorithm is moderately superior to other ML methods for predicting malnutrition in under-five children in Bangladesh.
• Recommends the application of RF classification, coupled with RF feature selection, for enhanced malnutrition prediction.