Students' stress prediction and explainable analysis based on improved decision trees.

Area of Science:

• Educational Psychology
• Computational Social Science
• Mental Health Informatics

Background:

• Student populations face significant academic, social, and career-related pressures.
• Accurate stress level prediction and identification of influencing factors are crucial for student well-being.

Purpose of the Study:

• To develop and validate an enhanced machine learning model for predicting student stress.
• To identify key factors contributing to student stress.

Main Methods:

• Compared nine machine learning algorithms to select an optimal base model.
• Enhanced a decision tree (DT) model using the Harris Hawks Optimization (HHO) algorithm.
• Applied Shapley Additive Explanations (SHAP) for model interpretation and feature analysis.

Main Results:

• The decision tree (DT) algorithm achieved a prediction accuracy of 0.909.
• The optimized Harris Hawks Optimization-DT (HHO-DT) model improved accuracy to 0.927 with fewer misclassifications.
• Shapley Additive Explanations (SHAP) identified blood pressure, social support, and depression as primary predictors of student stress.

Conclusions:

• The HHO-DT model offers a scientifically effective tool for predicting student stress.
• Findings support targeted interventions by educators, parents, and students to promote mental health.
• Accurate stress prediction can aid in developing strategies to alleviate student pressure and enhance overall well-being.