Development of a YOLO-Based Artificial Intelligence (AI) System for Early Detection of Stunting Risk in Children in 3T Regions of North Sumatra Province

Abstract

Stunting is a chronic nutritional problem that has long-term impacts on children’s physical growth, cognitive development, and future productivity. This condition remains a major challenge in the 3T regions (frontier, outermost, and disadvantaged areas) of North Sumatra Province due to limited healthcare personnel, lack of measurement facilities, and delays in early detection. This study aims to develop an artificial intelligence system integrating YOLOv8 and Random Forest to automatically and in real time detect stunting risk in children. The YOLOv8 model is utilized to detect the presence of a child and estimate height through visual image analysis, while the Random Forest algorithm classifies the risk level based on the Height-for-Age Z-score (HAZ) derived from anthropometric and demographic data. The dataset consists of 29 children from 3T regions, with training and testing splits used to evaluate model performance. The results show that the system achieved an accuracy of 97.8%, precision of 96.5%, recall of 95.9%, F1-score of 96.2%, and an area under the ROC curve (AUC) of 0.98. The system successfully detects children in real time, produces risk classifications consistent with manual measurements, and automatically documents examination data. The novelty of this research lies in the integration of YOLO for automatic height measurement and Random Forest for nutritional classification, which has not been applied in the 3T regional context. This system has the potential to serve as a digital tool for healthcare workers and posyandu cadres to accelerate child nutrition monitoring in an efficient, accurate, and well-documented manner.