Prioritizing at-risk bus drivers: a safety-constrained burnout severity classification model using information gain
Document Type
Journal Article
Publication Date
2026
Subject Area
ridership - drivers, planning - safety/accidents
Keywords
Traffic psychology, Driver burnout, Safety performance, Burnout severity classification, Information gain
Abstract
Job burnout is a critical occupational hazard that compromises the safety of public transport systems. Prevailing classification methods, however, often fail to establish a reliable, monotonic relationship between burnout severity and safety performance. To address this, we developed a novel framework that incorporates a safety-performance constraint, which requires the proportion of violation-involved drivers to increase with burnout severity. We constructed an information gain-based optimization model to identify the optimal burnout severity classification under this constraint. The framework was validated on a dataset of 1461 bus drivers, demonstrating its effectiveness. The model stratified drivers into four distinct tiers based on MBI-GS scores: no burnout [0, 0.87], mild (0.87, 2.27], moderate (2.27, 4.53], and severe (4.53, 6.00]. A clear, monotonic risk gradient was observed, with the proportion of drivers committing safety violations increasing consistently from 38.97 % (no burnout) to 46.26 % (mild), 51.40 % (moderate), and 60.00 % (severe). Comparative analyses confirmed the superiority of the proposed framework over conventional methods (Weighting and Dimensional Criteria). The framework achieved stronger correlations of the classified burnout levels with underlying burnout scores (r = 0.947 vs. 0.909 and 0.899) and safety outcomes (r = 0.131 vs. 0.093 and 0.088), higher information gain (IG = 8.6 × 10−3 vs. 4.3 × 10−3 and 3.9 × 10−3), and superior cluster validity (DBI = 0.4884 vs. 0.5693 and 0.9344). This indicates that, beyond most faithfully representing the continuum of burnout severity captured by the raw scores, the framework also enables a more precise characterization of violation risk. By translating burnout severity into a four-tiered risk classification with empirically defined violation rates (38.97 % to 60.00 %), this work provides transit agencies with a precise tool for identifying at-risk drivers and implementing targeted interventions, ultimately enhancing road safety.
Rights
Permission to publish the abstract has been given by Elsevier, copyright remains with them.
Recommended Citation
Wu, C., Chen, G., Gao, P., Zhang, S., He, R., & Liu, H. (2026). Prioritizing at-risk bus drivers: a safety-constrained burnout severity classification model using information gain. Transportation Research Part F: Traffic Psychology and Behaviour, 118, 103526.
Comments
Transportation Research Part F Home Page:
http://www.sciencedirect.com/science/journal/13698478