Mental stress among college students has escalated, driven by academic pressures and excessive internet use. This study evaluates stress levels during two critical phases: pre-examination week and internet-intensive periods. Chronic stress can lead to anxiety, depression, suicidal tendencies, and cardiovascular issues. To address this, an Enhanced RF AdaBoost Hybrid Classifier is proposed to predict and classify stress levels. Combining Random Forest’s feature selection with AdaBoost’s adaptive error reduction, the model achieves 96% accuracy—surpassing traditional classifiers like Decision Tree and SVM. Data were collected from diverse academic and demographic backgrounds, and evaluated using accuracy, precision, recall, and F1-score. The findings highlight a strong correlation between stress intensity and online activity, emphasizing the need for early detection tools. The RF-AdaBoost model proves effective for mental health surveillance and intervention, offering a reliable computational approach to support student well-being and reduce stigma around seeking professional help.