Deep Learning Models for Stress Analysis in University Students: A Sudoku-Based Study.

Area of Science:

• Psychophysiology
• Artificial Intelligence in Healthcare
• Educational Psychology

Background:

• University students face increasing academic and familial stress due to societal pressures like "involution."
• High stress levels negatively impact student well-being, necessitating effective monitoring methods.
• Existing stress detection methods often use artificial stimuli, which may not reflect real-world student stressors.

Purpose of the Study:

• To evaluate student stress levels using physiological signals during engaging, yet potentially stressful, tasks.
• To assess the efficacy of advanced AI models in accurately detecting stress in university students.
• To compare model performance under varying conditions, including distractions and social monitoring.

Main Methods:

• Collected physiological data (photoplethysmography - PPG, electrocardiogram - ECG, electroencephalogram - EEG) from students playing Sudoku.
• Employed enhanced deep learning models, including Long-term Recurrent Convolutional Networks (LRCN) and self-supervised Convolutional Neural Networks (CNN), for stress assessment.
• Validated model predictions against self-reported stress levels post-experimentation.

Main Results:

• The enhanced AI models demonstrated high proficiency in assessing student stress levels.
• Accuracy rates exceeding 95% were achieved, with F1-scores above 93%, particularly under distracting auditory conditions (95.13% accuracy, 93.72% F1-score).
• Near-perfect accuracy (98.78%) and high F1-score (95.39%) were observed under comforting conditions, and 97.76% accuracy with a 96.67% F1-score when observed by another individual.

Conclusions:

• The proposed AI models offer a robust and accurate method for monitoring student stress.
• Physiological signal analysis combined with advanced machine learning can effectively quantify stress in academic contexts.
• This approach provides a promising tool for educational institutions to support student mental health and well-being.