COMPARISON OF RANDOM FOREST, K-NEAREST NEIGHBOR, DECISION TREE, AND XGBOOST ALGORITHMS FOR DETECTING STUNTING IN TODDLERS

Stunting is a significant health issue in many developing countries, including Indonesia. Advances in health technology have opened new opportunities to improve the accuracy and efficiency of detecting stunting in young children, with one such advancement being Machine Learning technology. This study compares various Machine Learning algorithms for detecting stunting in children. The methodology includes data collection, data exploration, data preprocessing, feature extraction, model classification, and model evaluation. The results show that Random Forest demonstrates superior performance with the highest accuracy of 0.999132, recall of 0.999132, and a macro-averaged F1-score of 0.998906, making it the most consistent model for predicting child nutritional status. K-Nearest Neighbor also shows very good performance with an accuracy of 0.999050 and an F1-score of 0.998748. Decision Tree has an accuracy of 0.999091 and an F1-score of 0.998705, closely matching the performance of Random Forest and KNN. XGBoost, with an accuracy of 0.991033 and an F1-score of 0.987495, performs lower than the other three models. Therefore, Random Forest is the recommended choice for implementing stunting prediction in children.