Obesity is a growing global health issue, closely linked to serious conditions like type II diabetes, heart disease, and certain cancers. Early detection and intervention are essential for effective management and prevention. This study uses machine learning to classify obesity levels based on factors like diet, lifestyle, and physical characteristics. The dataset includes variables such as age, height, weight, family history, and physical activity. We compare two classification models: Random Forest and Logistic Regression, after preprocessing the data using label encoding and standardisation. Random Forest is selected for its capacity to handle complex, high-dimensional data, while Logistic Regression is chosen for its clarity and interpretability. We assess the models' performance using standard metrics like accuracy, confusion matrix, ROC curve, AUC, and the classification report. A multi-class approach is used to categorize individuals into normal weight, overweight, and various obesity levels, offering a more detailed picture than traditional BMI methods. The findings show that Random Forest significantly outperforms Logistic Regression, achieving 95.7% accuracy versus 87.4%. It also scores better on precision, recall, and F1-score. While Logistic Regression provides easier interpretation, it struggles with intermediate obesity levels. Random Forest’s robust performance and ability to pinpoint key factors in obesity make it a more effective tool for early detection and targeted interventions. Thus, we recommend Random Forest for practical healthcare applications, as it combines high accuracy with the potential for future improvements through advanced techniques like deep learning.