Behind the Black Box: Improving Stunting Determinants Analysis Through Explainable Artificial Intelligence

Abstract

Stunting is a public health problem that has a long-term impact on the quality of human resources. This study aims to analyze the performance of machine learning algorithms and identify the dominant factors of stunting using the Explainable Artificial Intelligence (XAI) approach. The dataset used was 120,999 toddlers with age, height, gender, and nutritional status attributes. The research stages include data pre-processing, normalization, separation of training and testing data (80:20), modeling using C4.5 algorithms, Support Vector Machine (SVM), and Random Forest, and evaluation using accuracy, precision, recall, and F1-score. The results showed that Random Forest and C4.5 achieved the best performance with an accuracy of 99.93%, while SVM achieved 98.37%. Interpretive analysis using SHAP revealed that height and age were the most dominant factors in the stunting classification with a contribution of 0.59 and 0.41, respectively, while gender contributed relatively small. These findings show that the integration of multi-algorithmic evaluation and XAI not only results in accurate prediction models, but also transparent and interpretive, thus supporting data-driven decision-making in efforts to accelerate stunting reduction in Indonesia