Analyzing Perceived Psychological and Social Stress of University Students: A Machine Learning Approach

Area of Science:

• Psychology
• Computer Science
• Public Health

Background:

• The COVID-19 pandemic significantly increased psychological and social stress among university students due to isolation, digital dependence, and reduced social activities.
• Early detection of student stress is vital for academic performance and mental well-being.
• Machine learning (ML) offers a promising approach for early stress prediction and intervention.

Purpose of the Study:

• To develop and validate a reliable machine learning-based prediction model for perceived stress in university students.
• To identify the prevalence of high social and psychological stress within the student population.
• To evaluate the effectiveness of different ML algorithms and feature reduction techniques for stress prediction.

Main Methods:

• A supervised machine learning approach was used, with data collected from 444 university students via an online survey.
• Feature reduction techniques included Principal Component Analysis (PCA) and the chi-squared test.
• Hyperparameter optimization (HPO) was performed using Grid Search Cross-Validation (GSCV) and Genetic Algorithm (GA).

Main Results:

• 11.26% of students reported high social stress, and 24.10% reported extremely high psychological stress.
• The developed ML models achieved high predictive performance, with the Multilayer Perceptron model combined with PCA and GSCV showing the highest accuracy (80.5%).
• Key performance metrics included accuracy (80.5%), precision (1.000), F1 score (0.890), and recall (0.826).

Conclusions:

• Machine learning models can effectively predict perceived stress levels in university students.
• The findings highlight a significant prevalence of psychological stress, necessitating targeted interventions.
• The study suggests potential for ML-driven strategies to support student well-being during stressful periods.