Exploring Stress Variation During Stressful Events by Using Electrocardiogram

This paper focuses on analyzing the variation of stress in real exam scenarios. We collected ECG data from 35 college students during an exam, divided the stress physiological data into three segments, i.e. early, middle and late durations of an exam, and used machine learning methods to built binary classification models of each pair of the above three exam durations, and obtained the best F1 scores of 84.21 %, 66.67% and 73.68%, respectively. Statistical test results showed that subjects had stronger stress levels during the early duration of the stressful exam than during the middle and late durations.