Machine learning techniques to predict overweight or obesity

Abstract

Overweight and obesity are considered a public health problem, as they are related to the risk of various diseases, and also to the risk of increased morbidity and mortality. The main objective of this work was to apply machine learning techniques for the development of a predictive model for the identification of people with obesity or overweight. The model developed was based on data related to the physical condition and eating habits. Furthermore, the machine learning classification algorithms that were tested were: decision tree,support vector machines, k-nearest neighbors, gaussian naive bayes, multilayer perceptron, random forest, gradient boosting, and extreme gradient boosting. Model hyperparameters were tuned to improve accuracy, resulting in that the model with the best performance was a random forest with 78% accuracy, 79% precision, 78% recall, and 78% F1-score. Finally, the potential of using machine learning models to identify people who are overweight or obese was demonstrated. The practical use of the model developed will allow specialists in the health area to use it as an advantage for decision-making.