Prediction of Obesity among Adults and Adolescents Using a Machine Learning Approach

Obesity has become one of the most significant public health issues of the 21st century. Obesity is a chronic, complex disease characterized by excessive fat accumulation that impairs health. It also has social and psychological dimensions, impacting individuals across all age groups and socioeconomic strata. It is associated with a range of risk factors, including diabetes, depression, and cancer. It presents a significant challenge to both developed and developing countries worldwide. This study aims to explore the factors contributing to obesity and develop a predictive model to identify individuals at risk. Using a secondary dataset, three machine learning algorithms were implemented for both interpretation and prediction. Logistic regression was used to identify significant associations between obesity and key factors such as family history of overweight and high caloric food intake. Among the predictive models, the random forest classifier demonstrated superior performance. The model was evaluated using metrics such as accuracy, specificity, and the Receiver Operating Characteristic (ROC) curve. The results indicated that the random forest classifier was the most effective model for predicting obesity, achieving the highest accuracy and ROC values. In conclusion, the findings of this study suggest that machine learning models, particularly the random forest classifier, can be effectively used to identify at-risk individuals and may offer valuable insights for the healthcare sector. The integration of such models could improve targeted interventions and support public health initiatives aimed at mitigating the obesity epidemic.