Machine Learning-Based Prediction of Psychological Well-Being Using Behavioral, Cognitive, and Digital Interaction Patterns

Cognitive, behavioral, demographic, and digital factors affect psychological well-being in university students, but their interaction in predicting psychological well-being is not fully characterized. This study aimed to develop and compare machine-learning models for predicting psychological well-being and identify the most influential predictors. Secondary cross-sectional data from undergraduate students aged 18–24 years were analyzed, with psychological well-being modeled as a continuous outcome using 17 behavioral, cognitive, demographic, and digital predictors. Linear Regression, Elastic Net, Random Forest, Gradient Boosting, and Support Vector Regression (SVR) were evaluated using nested five-fold cross-validation and an independent 20% test set. Model performance was assessed using R², RMSE, and MAE, while permutation importance quantified predictor contribution. The final analytical sample included 607 participants. SVR achieved the strongest independent test performance (R² = 0.4216, RMSE = 0.4216, MAE = 0.3015), with nested cross-validation showing closely comparable performance to Elastic Net. Growth mindset emerged as the most important predictor with a permutation R² importance of 0.9059 ± 0.1182, while the other variables of FOMO, social comparison, demographic characteristics and platform-use variables were not as important. The results indicate that psychological well-being can be accurately predicted at a moderate level by a comprehensible multivariable model, with the cognitive characteristics offering more predictive information than the general digital use metrics.