Quantifying Stress Levels in Night-Shift Bus Drivers: An Integrative Approach Utilizing Wearable Sensor Technology and Qualitative Evaluation

This research aims to create a real-time wearable sensor system for detecting stress level in night-shift bus drivers. Stress levels are assessed using motion/acceleration data and qualitative assessment. The driver's head-mounted accelerometer sensor system records motion data, which is then transmitted to an SD card for pre-processing. The study was conducted in real- life bus driving circumstances, and data was obtained from (n=12) subjects over a duration of 360±24 minutes, with a data sampling frequency of 100 Hz. A 10-point scale (1–10) questionnaire was used to measure privacy-related concerns and their impact on bus driver stress levels. To evaluate the dynamics of stress level, various classification methods, including Discriminant Analysis (DA), Decision Trees (DT), Support Vector Machine (SVM), K-Nearest Neighbour (KNN), and Naïve Bayes (NB) machine learning models, were employed. The Quadratic-SVM model performed the best, with an accuracy rate of 91.7% and an F-1 score of 92.8%; the Coarse Gaussian-SVM analysis performed the least well, with an accuracy of 82.0% and an F-1 score of 84.9%. In summary, this study presents a potential solution for categorizing night-shift bus driver stress into four distinct classes: no stress, low stress, medium stress, and high stress. It offers valuable insights for future researchers, addressing certainties in the domain of night shift bus driver stress indices and shedding light on trends to detect bus driver stress levels.