遊覽車客運業事故頻次之模化分析

Modeling Crash Frequencies of Tour Bus Companies

由於近年來全國遊覽車事故持續成長，故找出遊覽車事故之關鍵影響因素，俾利業者及主管機關據以改善至為重要。基此，本文蒐集遊覽車A1、A2事故及相關資料，利用頻次計數模式進行分析，並比較不同曝光量變數設定對模式推估之影響。結果顯示以離散性及mean absolute percentage error值而言，以延車公里為曝光量之零膨脹負二項迴歸模式表現最佳。依據該模式推估結果顯示：「違規件數」、「甲類大客車比例」、「駕駛平均年齡」、「女性駕駛比例」及「年資2年以下比例」為顯著變數，顯示遊覽車公司違規件數愈高、甲類大客車及年資較低之駕駛比例愈多、以及駕駛年齡愈高將導致事故件數提升，最後由模式推估結果提出因應對策。

The rising number of tour bus accidents highlights the urgent need to identify key contributing factors and develop effective countermeasures. This study employs count models to analyze crash counts among tour bus companies and assesses the impact of different exposure variables. The results indicate that the Zero-Inflated Negative Binomial model using bus-kilometers (bus-km) as the exposure variable performs best, based on dispersion tests and mean absolute percentage error. This suggests that bus- km is a more suitable measure of exposure for modeling crash frequency. Significant factors associated with higher crash counts include the number of violations, the proportion of large buses, the average age of drivers, the proportion of female drivers, and the proportion of junior drivers (with less than two years of experience). These findings imply that companies with more violations, a higher share of large or inexperienced drivers, and older drivers are more likely to experience crashes. Accordingly, targeted countermeasures are proposed to address these risk factors.