Enhancing Stunting Detection Accuracy in Children Using SVM with Advanced Data Balancing Techniques

Abstract

Stunting is a significant public health concern with enduring effects on children's physical and cognitive development. Traditional stunting classification methods often fail due to the complexity and imbalance of health data. This study proposes a novel technique combining Support Vector Machines (SVM) with Synthetic Minority Oversampling Technique (SMOTE) and Tomek Links to address these challenges. The proposed method was evaluated on a dataset from Kecapi Jepara, focusing on children's nutritional status before and after vitamin intervention. The results showed a significant improvement in classification accuracy, with an F1-score improvement of 12% and a 10% increase in overall accuracy compared to conventional methods. Specifically, the use of SMOTE and Tomek Links corrected the data imbalance, reducing the misclassification of stunted children by 15%. By incorporating these advanced machine learning techniques, the study offers a robust framework for early stunting detection, providing valuable insights for targeted public health interventions and contributing to global efforts to reduce stunting prevalence.