EUVIMEDEUVIMED entdecken
Crossref

Intelligent Optimization of Public Space Allocation in Urban Renewal Using Ensemble Learning

Bojuan Li

Transactions in GIS · 2026-11 · 30 (7)

DOI: 10.1111/tgis.70382 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Transactions in GIS, online erschienen am 09.11.2026. Verfasst von Bojuan Li. Die Arbeit nennt 46 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Urban renewal programmes require spatially differentiated public space allocation across sites that vary profoundly in thermal environment, land economy, and social structure–yet prevailing practice relies on non‐spatial, single‐criterion standards. We present a spatially‐aware stacking ensemble framework integrating four geographically referenced data streams–administrative project records, resident satisfaction surveys, temporal behavioral panels, and Sentinel‐2/VIIRS remote‐sensing composites–across 3500 renewal parcels in China (2018–2025). A Ridge‐meta stacking ensemble achieves composite R 2 = 0.875 under stratified five‐fold cross‐validation; external validity is tested by withholding an entire climate zone from training: applied without retraining to each of the five withheld zones in turn, the framework retains R 2 = 0.837–0.858, so transfer into a thermal regime absent from the training data costs at most 0.038 R 2 , while withholding arbitrary spatial blocks of sites instead costs almost nothing. In practical terms, the ensemble leaves no detectable geographic pattern in its errors: residual spatial autocorrelation is statistically indistinguishable from zero, and the optimism of random relative to spatially blocked cross‐validation is only 0.004 R 2 , so accuracy does not rest on spatial dependence between neighboring sites. SHAP analysis identifies urban heat island intensity as the dominant spatial predictor and reveals a place‐specific equity asymmetry: thermal optimization penalties fall disproportionately on low‐land‐value communities, motivating thermally‐indexed progressive subsidy policies. SMOTE augmentation recovers severely underrepresented Park allocations (F 1 : 0.414 → 0.634), exposing a structural spatial bias against ecologically critical space types. The framework provides planners with an interpretable decision‐support toolkit whose limits of transfer are quantified rather than assumed. Its two equity findings map directly onto the Sustainable Development Goals: the thermal penalty borne by low‐land‐value communities identifies exactly where SDG 10 (reduced inequalities) requires compensatory cooling investment, while the recovered Park allocations identify where SDG 11 (inclusive, safe and green urban space) requires minimum green‐space quotas.

Zitieren

APA

Li, B. (2026). Intelligent Optimization of Public Space Allocation in Urban Renewal Using Ensemble Learning. Transactions in GIS, 30(7). https://doi.org/10.1111/tgis.70382

Vancouver

Li B. Intelligent Optimization of Public Space Allocation in Urban Renewal Using Ensemble Learning. Transactions in GIS. 2026;30(7). doi:10.1111/tgis.70382

BibTeX
@article{Li2026,
  author = {Li, Bojuan},
  title = {Intelligent Optimization of Public Space Allocation in Urban Renewal Using Ensemble Learning},
  journal = {Transactions in GIS},
  year = {2026},
  volume = {30},
  number = {7},
  doi = {10.1111/tgis.70382}
}
Kostenlos starten →