【摘要】 交通出行行为预测研究是构建港珠澳三地智慧出行服务平台及提升交通管理水平的重要理论基础。首先，通过问卷调查方式采集了港珠澳大桥跨境出行样本数据；其次，针对样本数据集中所预测的类别数据标签并不均衡情况，引入BLSMOTE算法，构建多模式跨境出行行为模式预测模型；最后，利用SHAP方法分析特征变量对预测结果的影响。结果表明：BLSMOTE算法与随机森林模型相结合的BLSMOTE-RF模型与原始模型相比有更好的预测性能，在关于出行者的特征变量中，通行时间、是否进行港澳两地联游出行、安全性在出行行为模式选择中较为重要。该方法可用来预测跨境出行者多模式跨境出行情况，有利于向出行者提供个性化的跨境出行服务。更多还原

【Abstract】 Travel behavior prediction is an important theoretical basis for building smart cross-border travel systems and improving traffic management level in Hong Kong, Zhuhai and Macao. Firstly, this study collected cross-border travel sample data of Hong Kong-Zhuhai-Macao Bridge by questionnaire survey. Secondly, BLSMOTE algorithm is proposed to solve the problem of imbalanced category data labels predicted in sample dataset. Then the multi-mode cross-border travel behavior mode prediction model was built. Finally, the SHAP value is used to explain and analyze the influence characteristics of the prediction model. The results show that the hybrid model based on BLSMOTE algorithm and Random Forest model has better prediction performance than the original model. Among the characteristic variables of travelers, travel time, whether to take Hong Kong and Macao combined travel, and safety are more important in the choice of travel modes. The proposed method can be used to predict the multi-mode cross-border travel, which is conducive to providing personalized cross-border travel services to travelers.更多还原