Geospatial Integration for TOD

Comparing White Black and Grey‐Box Models of Population Density

Bibliographic Data

ID21652152
AuthorsLe Tung Duong (0009-0005-8250-0694, Faculty of Civil Engineering Ton Duc Thang University Ho Chi Minh City Viet Nam, corresponding author)
Year2026
Publication date2026-06-02
Peer ReviewedYes
Open AccessYes
TypeARTICLE
VenueSystems Research and Behavioral Science (JOURNAL)
Journal identifiersISSN: 1092-7026 • E-ISSN: 1099-1743
PublisherWiley (PUBLISHER • GB)
DOI10.1002/sres.70093
OpenAlexW7163155013
LanguageEN
References cited46

Transit‐oriented development (TOD) requires population‐density forecasts that are both accurate and interpretable. This study benchmarks three model families including white‐box (ordinary least squares), black‐box (nine ensemble/machine‐learning algorithms) and grey‐box (symbolic regression) within the Land Use Transport Feedback Cycle (LUTFC) framework for 1 km station buffers around 168 MRT stations in Singapore. The methodological contribution is a reproducible GIS workflow that integrates multi‐source geospatial data: URA Master Plan land‐use polygons; OpenStreetMap street/pedestrian networks and points of interest (POI); WorldPop population grids and the Built‐Settlement Growth Model; and ridership from LTA DataMall. From these layers, we compute LUTFC variables normalized by area (residential land share, network density, pedestrian crossing count, POI intensity, urban‐change ratio, ridership) and relate them to population density. Results reveal a clear accuracy–interpretability trade‐off: ensemble models achieve the highest predictive performance ( R 2 ≈ 0.66), linear regression attains R 2 ≈ 0.59 and symbolic regression approaches ensemble accuracy ( R 2 ≈ 0.64) while yielding closed‐form expressions that preserve interpretability. Across specifications, pedestrian crossing count and residential land share consistently dominate variable importance, reinforcing a pedestrian‐first TOD logic. The open, transferable GIS pipeline and grey‐box models together provide decision‐support that is both transparent and generalizable to rapidly urbanizing contexts where official data are limited or heterogeneous

Geographic information system · Geospatial analysis · Land use · Pedestrian · Population · Regression · Regression analysis · Workflow · Human Mobility and Location-Based Analysis · Transportation Planning and Optimization · Urban Transport and Accessibility

Tools

Open DOI