This study examines the correlation between physical exercise habits and mental health outcomes in adolescents using a regression-based Multilayer Perceptron (MLP) model. By utilizing a substantial dataset, the model successfully captures intricate, non-linear connections between several exercise-related characteristics and mental health scores. The MLP model exhibited robust prediction accuracy, characterized by elevated R² values and diminished MSE, underscoring its promise for practical implementations. By integrating big data, a thorough analysis was conducted, which demonstrated that engaging in regular physical activity is a substantial indicator of mental well-being. The results emphasize the significance of engaging in physical activity to enhance mental well-being and showcase the effectiveness of machine learning and big data in health evaluations. Subsequent research could investigate the integration of supplementary lifestyle elements and more sophisticated deep learning architectures to further improve prediction abilities.