Neural networks to estimate body weight using height, age and waist, in Peruvian population

Introduction: Estimation of body weight is essential in health assessments. Traditional methods can be limited by equipment availability or measurement biases, so the use of models based on artificial intelligence represents a promising alternative. Objective: To estimate body weight, using age, height, and abdominal circumference using multilayer perceptron neural networks. Methods: Cross-sectional analytical study carried out in 61,857 individuals from the 2022 and 2023 Demographic and Family Health Survey. Body weight, height, age, and abdominal circumference were analyzed. Multilayer perceptron neural networks and scatter plots with coefficient of determination (R²) were used. Results: Model training showed relative errors of 0.102 and 0.117 in the Demographic and Family Health Surveys of 2022 and 2023, respectively; in the test they were 0.099 and 0.112. Scatter plots indicated the model accounted for 90% and 89% of body weight variation in 2022 and 2023, respectively (R² of 0.899 and 0.885). Conclusions: The neural network based on height, age and abdominal circumference is an efficient tool for estimating body weight in the Peruvian population.