Public health mapping based on childrens nutritional status prediction using enhanced framework of machine learning

Public health monitoring involves community efforts to prevent disease, extend life, and promote health care, considering social, cultural, and economic influences. Surveillance process plays a crucial role in addressing health issues, with children's nutrition serving as a key indicator of community well-being. To identify the priority of intervention by the government, a map indicating the mapping of public health status is needed. Numerous previous studies have explored using machine learning models to analyze children's nutrition status. However, a definitive framework for predicting children_s nutritional status using machine learning remains uncertain, which is essential for creating a public health map. Developing and validating an enhanced machine learning framework for predictive analysis is necessary to generate this public health map. This study used various algorithms in the proposed enhanced framework, including Neural Networks, Random Forests, Decision Trees, Logistic Regression, and Extreme Gradient Boosting. The findings showed that the framework that employs a class imbalance handler offers better results. The Neural Networks method demonstrated the highest accuracy rates of 90.5% for wasting, 79.0% for underweight, and 74.8% for stunting, outperforming other methods. Performance comparisons are primarily accuracy-based but consider additional metrics like the F1 score for a comprehensive assessment. The framework for predicting WHZ (weight-for-height zscore) status combines Handler Class Imbalance Bagging (7:3) with Synthetic Minority Over-sampling Technique (SMOTE). Predicting HAZ (height-for-age z-score) status, the framework uses Handler Class Imbalance Bagging (7:3) augmented with Sample 300. The enhanced frameworks were employed to predict WHZ and HAZ status. Based on those predictions, the study proposed various public health maps, including clusterbased and threshold-based maps. This research implies the potential of the proposed machine learning frameworks to enhance public health intervention strategies by accurately predicting children's nutritional status and guiding targeted interventions.