Using Stress Indicators and AI to Predict Student Academic Performance

Stress significantly affects students’ academic performance and mental health, impairing cognitive functions such as memory, attention, and problem-solving. Chronic stress due to academic pressures can lead to long-term consequences for well-being and educational outcomes. This study explores the link between physiological stress indicators and academic performance by using vital signs to predict exam scores and identify at-risk students. Integrating wearable technology with artificial intelligence enables proactive stress management and timely interventions. Data were collected from a PhysioNet cohort involving ten students who wore devices to monitor heart rate (HR), blood volume pulse (BVP), and body temperature (TEM) before, during, and after exams. From ۱,۲۰۰ data points per student, ۴۰ were randomly selected for unbiased sampling. Two machine learning models—Multilayer Perceptron (MLP) and Decision Tree—were used to predict exam scores. The dataset was split into ۸۰% training and ۲۰% validation. The MLP model outperformed the Decision Tree, achieving an R-squared of ۰.۹۳, MSE of ۴۸.۳۹, and RMSE of ۶.۹۵, showing strong predictive accuracy. The Decision Tree provided interpretable thresholds for physiological changes, complementing MLP’s results. Findings suggest that AI-driven analysis of physiological signals can effectively identify students facing stress-related challenges. Combining wearables with machine learning allows real-time monitoring and targeted support, such as counseling or mindfulness programs, improving both academic success and mental resilience. While the small sample size limits generalizability, this study supports AI-based systems in education to reduce stress effects and enhance student well-being. Future research should expand datasets and include additional factors like study habits.

کلیدواژه ها:

Stress Indicators ، Academic Performance ، Machine Learning Models