Skip to content
Conference Open access

SLP-LR: A Predictive Analytics Approach for Mental Stress Detection among University Students

Sep 2026 · 2026 10th International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC) · pp. 2192-2197 · 0 citations · 15 references

Abstract

Student stress can adversely affect academic performance and mental well-being, creating a need for scalable support tools in higher education. This study investigates multiclass student stress classification using a dataset of 2,028 Bangladeshi university-student responses containing demographic variables and ten academic-stress questionnaire items. The dataset includes Low Stress (115 samples), Moderate Stress (1,348 samples), and High Stress (565 samples) categories. An 80/20 stratified split was used, and Synthetic Minority Oversampling Technique (SMOTE) was applied to the training data to address class imbalance. Random Forest, XGBoost, and multinomial Logistic Regression were evaluated. On the original 406-sample hold-out test set, Random Forest achieved 95.32% accuracy, XGBoost achieved 94.83% accuracy, and Logistic Regression achieved the highest accuracy of 98.77%, with macro-F1 and weighted-F1 scores of 0.9786 and 0.9878, respectively. A Streamlit-based prototype was developed for interactive prediction, probability visualization, and PDF report generation. Since the stress label is questionnaire-derived and Q1–Q10 are included as model inputs, the results should be interpreted as internal classification performance on the available questionnaire-based dataset. Leakage-controlled validation, repeated cross-validation, and independent institutional validation are identified as future work.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.