<![CDATA[This study applies a quantitative approach based on computational experiments to develop an accurate student stress level prediction model. The research design employs a cross-sectional method with data collection via online questionnaire instruments integrating multidimensional variables, including the Perceived Stress Scale (PSS-10) as the target variable, as well as the Pittsburgh Sleep Quality Index (PSQI) and Self-Compassion Scale (SCS) as key predictor features alongside academic variables. The main challenge of imbalanced data is addressed by applying the Synthetic Minority Over-sampling Technique (SMOTE) during the preprocessing stage to synthesize minority class samples and prevent majority bias. In the modeling phase, the Random Forest Classifier algorithm is utilized due to its superiority in handling complex non-linear relationships, and its performance is compared with Logistic Regression as a baseline model. Model validation is conducted using the 10-Fold Cross-Validation method to test data generalization. Performance evaluation focuses on Recall, Precision, and F1-Score metrics to ensure the model's sensitivity in effectively detecting high-stress cases as a clinically relevant early warning system.]]>
Copyrights © 2025
