| With the development of social economy and the increase of residents’ demand for intercity travel,it is increasingly important to build a high-quality intermodal travel system for passengers.Because the comprehensive transportation hubs are the carrier of intermodal travel,the coordination and cooperation among hub agglomerations are more and more critical.Airrail intermodal travel(ARIT)is a typical intermodal travel mode,which helps to expand the coverage of transportation hubs and provide rich intercity travel options.The choice of transfer cities is a key process of air-rail intermodal travel because different transfer cities will directly cause differences in the economy,speed and convenience of intermodal services.However,the ARIT service in China is still in its early stages: research covering a wide range of hub cluster scenarios is lacking,the ARIT passengers’ preferences and the inter-hub interactions are unclear,and the quantitative exploration of relevant factors is also insufficient.Therefore,this paper relied on the National Key Research and Development Program(2018YFB1601300),the National Natural Science Foundation of China(52072066),and the Jiangsu Provincial Outstanding Youth Fund Project(BK20200014),to carry out research on the key process of ARIT as the choice of transfer cities in transportation hub agglomerations.First,this paper applied a hierarchical clustering algorithm to evaluate the ARIT transfer services of the cities in two hub agglomerations,Beijing-Tianjin-Hebei and Yangtze River Delta,thereby analyzing the differences in their infrastructure constructions.Second,based on the desensitization data of Internet users’ historical ticket purchase orders,a descriptive statistical analysis was carried out on the spatial distribution,itinerary attributes and personal characteristics of ARIT passengers in these two hub agglomerations.Then,integrating the theory of discrete choice models,and reconstructing the modeling process through complementary of alternatives’ attributes,dataset reorganization and soft classification,an ensemble machine learning algorithm that extreme gradient boosting(XGBoost)was implemented to model ARIT passengers’ choice of transfer city.Meanwhile,several interpretable machine learning methods,such as permutation feature importance,SHAP value,and accumulated local effects,were used to identify feature importance and the complex nonlinear and interactive effects of various attributes on passengers’ choice of transfer city behavior in the ARIT scenario.Finally,focusing on the two routes,Baoding-Shanghai: a hinterland city of BeijingTianjin-Hebei to a metropolis,and Suzhou-Beijing: a prefecture-level city in the Yangtze River Delta to the capital,sensitivity analysis was conducted for different kinds of cities in terms of discount fare,increment of train speed,schedule settings,and transfer connectivity,to quantify the impact of corresponding service improvement strategies on the ARIT passenger flow in transportation hub agglomerations.The results show that 1)the XGBoost model achieves better performance in predicting ARIT transfer city choice than the conventional discrete choice model and can efficiently deal with the unbalanced dataset;2)attributes related to ARIT services in hub agglomerations are more important than individual demographic characteristics,and service-related attributes at the economic and efficiency aspects have nonlinear impacts;3)passengers from the two hub agglomerations,Beijing-Tianjin-Hebei and Yangtze River Delta,show a certain degree of heterogeneity when choosing ARIT transfer cities,and there are differences in the inter-hub collaborative relationship within different hub agglomerations.Then,this study provided relevant policy suggestions for promoting the development of air-rail intermodal travel in hub agglomerations. |
PDF Full Text Request