北京建筑大学学报2025，Vol.41Issue(3)：66-77,12.DOI:10.19740/j.2096-9872.2025.03.08
基于可解释性机器学习的公交出行选择行为影响因素研究
Research on the Influence Factors of Bus Travel Behavior Based on Explainable Machine Learning
摘要
Abstract
In order to explore the mechanism behind residents'choice of bus travel,a method that combines adaptive synthetic sampling(ADASYN)and the XGBoost(Extreme Gradient Boosting)model with Bayesian parameter optimization is proposed.The model is established based on the data of residents'travel behavior survey and public transportation travel intention survey within a week,and the SHAP attribution analysis(Shapley Additive Explanation)is used to explain the mechanism of the model.The results show that the XGBoost model with ADASYN oversampling and Bayesian hyperparameter optimization performs well with an accuracy of 90.48%and the area under the curve reaches 0.91;city category,education background,purpose of travel and the influence of other modes of transportation are all important factors for residents to decide whether to take bus or not.Among them,the type of city is the most important influencing factor.In different types of cities,travel distance,waiting time and the competitive factors of other modes of transportation have different effects on the result.In large cities,travel distance between 5 to 10 km has a positive effect on public transportation,while in supercity and megacity the effect remains negetive.Only in large cities,waiting time of more than 25 minutes will inhibit residents from taking public transportation,in megacity and supercity the effect behaves vaguely.In general,walking,bicycles,shared(electric)bicycles,subways and other modes of transportation are in a complementary cooperative relationship with public transportation,while private cars are in a mutually exclusive competitive relationship.The specific form of action varies with the scale of the city and the spatiotemporal characteristics of travel.关键词
交通工程/出行行为选择/可解释性机器学习/竞合关系/交互效应Key words
traffic engineering/travel mode choice/explainable machine learning/coopetition/interaction分类
建筑与水利引用本文复制引用
焦朋朋,陈世杰,罗薇,王健宇,唐玉洁..基于可解释性机器学习的公交出行选择行为影响因素研究[J].北京建筑大学学报,2025,41(3):66-77,12.基金项目
国家自然科学基金项目(52172301) （52172301）
北京市西城区优秀人才培养资助项目(202338) （202338）
北京建筑大学研究生创新项目(02081024003). （02081024003）
