Estimation of Weight in Children Under Five Years of Age Using Height and Age via Multilayer Perceptron Neural Networks

Guevara Tirado A. · Revista Cubana de Pediatria (2026)

Resumen

Introduction: Body weight in children is crucial for nutritional assessments and the evaluation of normal development. Objective: To estimate body weight in children under five years of age based on height and age using multilayer perceptron neural networks. Methods: An analytical, cross-sectional study was conducted using data from the 2022 and 2023 Demographic and Family Health Surveys. The study population consisted of 21,995 children under five years of age in 2022, and 21,111 in 2023. Multilayer perceptron neural networks and scatter plots were employed to calculate the coefficient of determination (R2). Results: In 2022, the multilayer perceptron yielded a coefficient of determination (R2) of 0.893 between the predicted body weight values and the values measured via scale; this indicates that 89% of the values predicted by the perceptron aligned with the actual body weight of children under five. In 2023, the coefficient of determination (R2) was 0.902, meaning that 90% of the values predicted by the perceptron aligned with the actual body weight of children under five. The model demonstrated high precision and reliability, exhibiting minimal differences between estimated and actual weights, acceptable margins of error, and a strong correlation (0.960 in 2022; and 0.958 in 2023). Conclusions: The application of multilayer perceptron neural networks enabled the estimation of body weight based on height and age in children under five years of age within the Peruvian population.

Instituciones: Universidad Científica del Sur