Comparison of Deep Neural Networks and Random Forest Algorithms for Multiclass Stunting Prediction in Toddlers

Abstract

Stunting in toddlers is a serious global health issue, with long-term impacts on physical growth and cognitive development. To address this problem more effectively, it is crucial not only to identify whether a child is stunted but also to predict the severity of the condition. Multiclass stunting prediction offers deeper insights into a child’s condition, enabling more precise and targeted interventions. This study aims to compare the performance of multiclass stunting prediction models using two machine learning algorithms: Deep Neural Networks and Random Forest. The research process involved data collection, preprocessing, as well as model development and testing. The results show that the Random Forest model achieved 100% accuracy in training and 99.92% accuracy in testing, while the Deep Neural Networks model achieved 93.49% accuracy in training and 93.21% in testing. Both models demonstrated strong performance in multiclass stunting prediction, with Random Forest proving superior in terms of accuracy compared to Deep Neural Networks.