Predicting Obesity Levels Based on Lifestyle and Activity Patterns

Abstract

Obesity is a growing global health concern driven by genetic, behavioral, and environmental factors. Machine learning (ML) offers potential for predicting and classifying obesity, however data accessibility and model scalability present challenges. This study evaluates various machine learning algorithms for obesity prediction, including Random Forest (RF), K-Nearest Neighbors (KNN), Gradient Boosting, and Support Vector Machines (SVM). The dataset comprises 1,610 individuals, considering health, behavioral, and demographic characteristics. The aforementioned metrics were employed to evaluate the model performances; accuracy, precision, recall, and F1-score. Of interest, Logistic Regression had the lowest accuracy score (76.39%), while Gradient Boosting had the highest 
(88.82%). Similarly, Gradient Boosting performed well with the other metrics, reinforcing its valid conclusions for obesity classification. There is substantial potential for machine learning, as demonstrated in this study, as it will enable the early detection of obesity, and provide intervention prospects for health management professionals.