Social media is increasingly embedded in students’ daily lives, influencing their attention, emotional regulation, and academic focus. Yet, most research examines academic success, social media use, and mental health separately. This study proposes a dual-objective machine learning framework that simultaneously predicts academic success and negative social media impact among higher education students. Two datasets from an institution were used: pre-admission records comprising demographic details, interview scores, and subject marks, and a questionnaire assessing social media interference with sleep, mood, and stress. Binary labels were defined for academic success (Successful/Not Successful) and social media impact (Negative Impact/No Impact). After label encoding and median imputation, five classifiers—Random Forest, Support Vector Machine, Multi-layer Perceptron, Gradient Boosting, and a soft-voting ensemble—were trained using an 80/20 split. Gradient Boosting achieved perfect test accuracy in both predictive tasks, followed by the ensemble and Random Forest models with similarly high accuracy. The Support Vector Machine and neural network performed moderately but showed weak detection of minority outcomes. The study underscores the effectiveness of gradient-boosted ensembles for educational data mining and demonstrates how integrating behavioral and mental-health indicators with academic data can enhance AI-enabled systems for student performance analytics and institutional decision-making.