Comparative Analysis of Machine Learning Algorithm Performance in Predicting Stunting in Toddlers

Abstract

Stunting is a condition where the growth of children and toddlers is stunted, which causes children to be shorter than they should be. In the long term, stunting can reduce reproductive health, study concentration, and work productivity, thereby causing significant state losses. The prevalence of stunting in Indonesia, which is still above 20 percent, shows that there are still chronic nutritional problems among toddlers. To prevent this from happening, identification as early as possible can be done using machine learning for predictions. The aim of this research is to conduct a comparative analysis of the performance of machine learning algorithms for predicting stunting in toddlers. Random Forest, K-Nearest Neighbors, and Extreme Gradient Boosting are the algorithms that are compared for their performance. The performance of each algorithm is measured using evaluation matrices such as accuracy, precision, recall, and f1-score. The research method starts with data collection, data preprocessing, data splitting, application of machine learning algorithms, evaluation of algorithm performance, and comparison of results. The performance evaluation matrix measurement results show that Random Forest has an accuracy of 99.95%, precision of 99.89%, recall of 99.94%, and f1-score of 99.91%. K-Nearest Neighbors has an accuracy of 99.93%, precision of 99.87%, recall of 99.88%, and f1-score of 99.88%. Meanwhile, Extreme Gradient Boosting has an accuracy of 99.36%, precision of 98.86%, recall of 98.95%, and f1-score of 98.90%. From the results of all performance evaluation matrices, it can be concluded that the random forest algorithm is the best algorithm for predicting stunting in toddlers.