Deciphering the influence of TOD on metro ridership: An integrated approach of extended node-place model and interpretable machine learning with planning implications. (October 2022)

• Record Type: Journal Article
• Authors: Su, Shiliang; Wang, Zhuolun; Li, Bozhao; Kang, Mengjun
• Is Part Of: Journal of transport geography. Volume 104(2022)
• Publication Date: 2022-10
• Subjects: Transit-oriented development -- Metro ridership -- Random forest regression -- Shapley additive explanations -- Nonlinear association; Transportation -- Periodicals; Telecommunication -- Periodicals; Transport -- Périodiques; Télécommunications -- Périodiques; Telecommunication; Transportation; Periodicals; 388
• DOI: 10.1016/j.jtrangeo.2022.103455
• Languages: English
• ISSNs: 0966-6923

Abstract

Abstract: Many global high-density cities have embraced transit-oriented development (TOD) strategies around metro stations in a strong push toward promoting transit trips. However, the general TOD principles must be adapted to a variety of local features. Thus, conducting a baseline study that deciphers the locality-specific differences and similarities in the influence of TOD on metro ridership is of paramount importance. Using the case of Shanghai in China, this paper demonstrated a novel approach to unveiling the locality-specific influence of TOD on metro ridership that integrated the node-functionality-place model with interpretable machine learning. By exploring a wide range of explanatory variables, we discovered that the relative importance of TOD structural factors (e.g., node, place and functionality) and neighborhood sociodemographics as well as their interactions presented great spatial and temporal heterogeneity for boarding and alighting on weekdays and weekends. Among the TOD structural factors, functionality was of the highest importance, and a greater contribution was observed within the central districts during peak hours. Regarding the interactions among the TOD structural factors, the interaction between node and functionality presented the highest relative importance, followed by that between place between functionality and that between node and place. Additionally, neighborhood sociodemographics also accounted for a noticeable contribution, especially in the outskirts. Based on the locality-specific estimations, the affinity propagation clustering algorithm was further used to cluster the TODs into difference groups as nuanced representations of the TOD-metro ridership relationships. The findings refreshed the knowledge base concerning the spatiotemporal heterogeneity in the nonlinear influence of TOD on metro ridership and provided new insights into spatial planning. The proposed approach showed high computational feasibility with strong theoretical underpinnings and thus offered broad potential applications into other high-density cities worldwide.