Predicting exam performance from stress signals using ML
New arXiv paper tests whether physiological markers recorded during exams—heart rate, skin conductance, temperature—can forecast academic outcomes.
• Compared logistic regression, random forest, SVM against transformers, LSTM, GRU
• Stress indicators show measurable correlation with exam results
• Transformers adapted best to complex signal interactions