Understanding the determinants of psychological well-being is crucial for promoting mental health. However, studying well-being presents challenges, including identifying clear definitions and valid measures, which Machine Learning (ML) could help to overcome. This study employed Random Forest Regression (RFR), Support Vector Regression (SVR), XGBoost (XGB), and Linear Regression (LR) to explore predictive features associated with psychological well-being in the HCP Young Adult cohort. By integrating feature importance analysis and explainable approach, SHapley Additive exPlanations (SHAP), we identified key predictors of psychological well-being, focusing on the Emotional Battery of the National Institute of Health (NIH) Toolbox. Both LR and RFR demonstrated similar performance in terms of errors, with RFR exhibiting a slightly lower mean absolute error (MAE) of 3.8 compared to Linear Regression’s MAE of 3.8. However, the Linear Regression model demonstrates a slightly higher R2, of 0.481, compared to 0.47 for RFR. SVR and XGB showed lower predictive power, with higher MAE values (5.3 and 5.5). The feature importance and SHAP showed that models shared the first three most important variables, although in a slightly different order: sadness, emotional support and perceived stress. Notably, according to SHAP, sadness emerged as the first predictive feature across models, except for SVR. In conclusion, our study underscores the potential of ML to serve cognitive sciences and enhance our comprehension of psychological well-being. By employing advanced ML techniques, we can glean valuable insights and develop targeted interventions for promoting mental health.