Explainable artificial intelligence for predictive modeling of student stress in higher education

Area of Science:

• Educational Psychology
• Computational Social Science
• Health Informatics

Background:

• Student stress is a significant issue in higher education.
• Current detection methods are often expensive and lack transparency.
• There is a need for affordable, scalable, and interpretable stress management tools.

Purpose of the Study:

• To develop a cost-effective, survey-based stress classification model for university students.
• To utilize multiple machine learning algorithms and eXplainable Artificial Intelligence (XAI).
• To support transparent and actionable decision-making for educational institutions.

Main Methods:

• Applied a supervised machine learning pipeline to student survey data.
• Trained and optimized six classification algorithms (Logistic Regression, SVM, Decision Tree, Random Forest, Gradient Boosting, XGBoost).
• Used SHAP analysis for model interpretability and feature importance ranking.

Main Results:

• Random Forest achieved the highest accuracy (0.89), followed by XGBoost (0.87).
• Key predictors of student stress identified: blood pressure, perceived safety, sleep quality, teacher-student relationship, and extracurricular activity participation.
• Both physiological and psychosocial factors significantly contribute to stress prediction.

Conclusions:

• A survey-based machine learning approach is effective for classifying student stress.
• Interventions addressing health, safety, support, relationships, and engagement can mitigate student stress.
• Findings support the development of integrated policies for enhanced student well-being in higher education.