Child Stunting Classification using the LightGBM Method: A Case Study in the Rowosari District of Kendal, Central Java

Stunting in children under five is a major health problem in Indonesia, significantly affecting physical, mental, and cognitive development, which can ultimately lower their quality of life in the future. Early identification of children at risk of stunting remains a challenge due to limited resources and the effectiveness of existing prediction methods. Machine learning techniques offer a promising approach to improving stunting classification and risk prediction. This study aims to develop an accurate and efficient classification model for identifying stunting status using the Light Gradient Boosting Machine (LightGBM) method. The study was conducted on children in Rowosari District, Kendal, Central Java, utilizing seven key variables: gender, age (months), birth weight, birth length, current weight, current height, and stunting status. The results indicate that the LightGBM model achieved 97% accuracy with an AUC value of 0.99, demonstrating a high ability to distinguish stunting status. Furthermore, the model successfully identified the key risk factors contributing to stunting. These results show that the LightGBM method has the potential to make very accurate predictions and also help people come up with more timely and targeted intervention strategies that are based on data. By leveraging machine learning, health practitioners and policymakers can improve stunting prevention efforts, optimize resource allocation, and implement more effective public health strategies to reduce stunting prevalence in Indonesia.