Obesity is a critical global health challenge, characterized by its complex etiology and association with numerous chronic diseases. Leveraging machine learning (ML) techniques offers promising avenues for improving obesity classification and risk prediction. This study aims to evaluate the efficacy of various ML algorithms, including Decision Trees (DT), Extra Trees Classifier (ETC), Multilayer Perceptron (MLP), Random Forest (RF), and Support Vector Machines (SVM), combined with diverse sampling techniques to address class imbalance. The research utilizes the publicly available Obesity Dataset, encompassing demographic and lifestyle variables. A stratified k-fold cross-validation approach was employed for robust model evaluation, and data balancing methods such as SMOTE and SVMSMOTE were implemented to enhance classification performance. Among the evaluated models, ETC demonstrated the highest accuracy (91.93%) and AUC (97.99%) when paired with SMOTE, underscoring its potential for scalable and precise obesity classification. These findings highlight the importance of integrating advanced ML methods and sampling strategies to tackle class imbalance. In addition, this study provides an important basis for the development of more effective decision-support systems in public health and clinical applications and paves the way for innovative approaches in the fight against obesity.