Abstract: To investigate the differences in driving behavior between cars and trucks in highway merging areas and improve the accuracy of risk prediction, this study analyzes the spatiotemporal trajectories and operational characteristics of different vehicle types in merging zones based on the HighD highway trajectory dataset. To further capture behavioral variations, ten microscopic driving features were extracted. After applying factor analysis for dimensionality reduction, the K-means clustering algorithm is employed to identify distinct driving styles. The study further incorporated driving style features to construct risk prediction models based on support vector machine (SVM), decision tree (DT), AdaBoost, and XGBoost. The results indicate that different vehicle types exhibit significant differences in speed and acceleration distributions in merging areas; the classification of driving styles can effectively characterize behavioral heterogeneity, and the incorporation of driving styles significantly improves risk prediction performance, among which the XGBoost model performs the best. This study provides a rigorous theoretical basis and practical insights for traffic safety modeling and personalized early-warning strategies in highway merging areas.

Key words: highway merging area; traffic characteristics; factor analysis; driving style; K-means clustering; risk prediction

摘要： 为探究汽车与卡车在高速公路合流区的驾驶行为差异并提升风险预测精度，研究基于 HighD 高速公路轨迹数据，分析了不同车型在合流区的时空轨迹与运行参数特性。为进一步 揭示驾驶行为差异，提取 10 项微观运动特征指标，运用因子分析法进行降维处理，并结合 K-means 聚类方法识别驾驶风格。研究进一步引入驾驶风格特征，构建基于支持向量机 （SVM）、决策树（DT）、AdaBoost 和 XGBoost 的风险预测模型。结果表明，不同车型在合 流区域表现出显著的速度与加速度分布差异；驾驶风格划分能够有效表征驾驶人行为特征， 引入驾驶风格后对风险水平预测有显著提升作用，其中 XGBoost 模型表现最优。该研究为高 速公路合流区的交通安全建模与个性化预警策略提供了理论基础与实践参考。

关键词: 高速公路合流区；交通特性；因子分析；驾驶风格；K-means 聚类；风险预测