An SVM-based identification of acute stress in pilots using objective indicators

Sudden incidents during flight trigger acute stress in pilots, compromising safety. An approach task was designed, with an engine failure during the landing phase to induce acute stress in 37 flight cadets using a C172-G1000 simulator. State anxiety scores, heart rates, heart rate variability, salivary cortisol concentrations, flight altitude, and heading were collected. Results revealed significant differences in physiological, biochemical, and behavioural data between stress and non-stress states. A support vector machine model was trained through feature selection, normalisation, and hyperparameter tuning. The model achieved an accuracy of 86.49% in distinguishing stress from non-stress states. This study provides a methodology for objective monitoring of pilot stress levels.