Quantifying internal and external heterogeneous effects of bus drivers' personality traits and driving behavior on...

Existing studies about bus accident analysis have examined the impacts of internal or external heterogeneity from multiple perspectives, highlighting the multifaceted nature of accident risk factors. However, there are core limitations still to be addressed: reliance on single-type physiological, psychological or driving behavior data, neglecting group classification stability in accident heterogeneity analysis, and insufficient guidance for bus on-board Advanced Driver Assistance System (ADAS) and Driver State Monitoring System (DSM) applications, resulting in incomplete crash mechanism interpretation, reduced intervention specificity, and inadequate tailored monitoring system adaptability. To address these gaps, a hybrid analytical framework (K-means-RPBLM) based on multi-source operational data and driver data was proposed in this study to quantify the internal and external heterogeneous impacts of personality traits and driving behavior on bus crashes. First, bus drivers were categorized into Robust, Ordinary, and Anxious personality trait groups through the K-means clustering algorithm. Subsequently, random parameter binary logit models (RPBLM) were estimated for the overall bus driver sample and each personality subgroup, and pairwise Wald tests were conducted for selected cross-group coefficient comparisons. The results revealed distinct subgroup-specific association and random-parameter patterns. Pairwise Wald tests showed significant cross-group coefficient differences for 20-30 years of driving experience across all three pairwise comparisons and for speeding-warning frequency between the Robust group and the other two groups, whereas no significant cross-group difference was detected for distracted-driving warnings. These findings provide partial, factor-specific evidence of personality-related heterogeneity in crash-risk mechanisms and support differentiated safety management for bus drivers.