Deteksi Dini Stunting pada Balita Menggunakan 1D Convolutional Neural Network (1D-CNN) pada Data Antropometri Numerik

Abstract

Stunting remains a major public health challenge in Indonesia, with a national prevalence of 21.6%. Its impact extends beyond impaired physical growth to affect cognitive development and long-term productivity. Early detection is typically performed through manual anthropometric measurements and Z-score calculations, which are relatively impractical and prone to computational errors, especially in resource limited settings. This study proposes a one-dimensional convolutional neural network (1D-CNN) based approach to detect stunting in children under five using numerical anthropometric data of age, sex, and height without manual feature engineering. The model was evaluated on 120,999 samples and achieved a recall of 99.3%, with only 4 out of 552 stunting cases going undetected, demonstrating strong ability to minimize false negatives in the context of public health screening. In comparison, the Random Forest model achieved 99.9% accuracy and an F1-score of 98.2%, demonstrating excellent overall classification performance. Nevertheless, 1D-CNN offers architectural advantages through automatic representation learning based on one-dimensional signal structures, making it more adaptable to the inclusion of sequential variables, the integration of longitudinal growth sensor data, and the development of future IoT based monitoring systems. Therefore, the proposed approach is not only competitive in detection performance but also provides greater scalability and flexibility for the continued development of digital screening systems at the primary healthcare level.