Classification of Stunting Cases Using Support Vector Machine and Random Forest Methods

Abstract

Stunting remains a serious public health challenge in Indonesia, severely limiting children’s physical and cognitive development. In Depok City, local health efforts like those in Cinere District actively aim toward a “Zero Stunting” target, although field monitoring still requires more adaptive and systematic screening instruments to process monthly anthropometric data. This work presents a preliminary comparative experiment testing two basline machine learning algorithms, Support Vector Machine (SVM) and Random Forest, to categorize specific stunting severity categories using localized health center data. The dataset consists of 101 historical patient records from UPTD Puskesmas Cinere, utilizing age, gender, weight, height, and the Height for Age Z-score (HAZ). Data preprocessing, segmentation, and assessment were performed in Pythin using Pandas and Scikir Learn modules. The experimental results demonstrate that the Random Forest model attained a moderate overall accuracy of 61.90%, while the SVM model reached 57.14%. Although the SVM showed superior threshold-level separation with an Area Under the Curve (AUC) of 0.76 in contrast to the Random Forest 0.63, both models reveal distinct performance limitations stemming from significant minority class disparities. These initial findings indicate that although baseline classifiers provide valuable diagnostic information, additional methodological advancements are essential before these models can be reliably utilized as trustworthy decision support systems in field screening settings.