.png){kind=icon}
.png){kind=icon}
Return
UAM vertiport location selection using XGBoost-based GIS data analysis: A Seoul-Daejeon case study
.png){kind=icon}
OA
AI
U
J
D
K
DOI：10.1080/12265934.2026.2691194.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Urban Air Mobility (UAM) is emerging as a next-generation urban mobility solution with the potential to alleviate congestion and enhance overall mobility efficiency. One of the critical challenges in UAM is determining optimal locations for vertiports. In this study, we propose a novel approach utilizing geospatial data processing and machine learning (XGBoost) for UAM vertiport location selection. 15 social, economic, transportation, and safety-related variables were collected and processed into a 1km × 1km grid. The model was trained using 4 expert-selected vertiport locations in Seoul (Gimpo Airport Vertiport, Yeouido Park Heliport, Jamsil Han River Park, and Suseo Station) and then applied to Daejeon, to assess its generalizability. To evaluate the consistency between the XGBoost model prediction and expert selection, two validation methods were employed: the Top-N Hit Rate and Rank-Biased Overlap (RBO). As a result, XGBoost achieved robust prediction performance, with SHAP analysis indicating height and accessibility to transportation infrastructures as the influential factors. The Top-10 Hit Rate indicated substantial overlap between the XGBoost model prediction and expert selection, while the RBO analysis revealed moderate agreement in ranking structure, particularly among high-ranked candidate locations. The model successfully identified vertiport candidate locations in Daejeon that align with expert intuition, demonstrating both accuracy and versatility. Our findings demonstrate the feasibility of machine learning-based vertiport location selection and provide insights into urban air mobility infrastructure planning.
Keywords:
Urban air mobility (UAM)
vertiport
machine learning
XGBoost
geographic information systems (GIS)
Journal
I
IF:
3
Papers:
463
Citations:
1.1K
Organization
Cited Papers