Clustering-based Framework for Large Urban Studies: A Case Study in Taipei City
Part Of
Building Simulation Conference Proceedings
Journal Volume
19
ISSN
25222708
ISBN (of the container)
9781775052043
ISBN
9781775052043
Date Issued
2025
Author(s)
Abstract
Urban morphology complicates large-scale studies, while urban grouping methods remain fragmented. To address this gap, this study developed a GIS-based clustering workflow to classify urban blocks in Taipei using eight key parameters. K-Means, DBSCAN, Mean-Shift, and Spectral Clustering are applied, with performance evaluated using Davies-Bouldin Index and Silhouette Score. Results indicate that K-Means (k = 4) and Spectral Clustering (n = 5) generate well-distributed clusters, while density-based methods produce biased results. K-Means effectively analyzes numerical attributes whereas Spectral Clustering better captures categorical features. GIS visualization enhances interpretation, offering insights for urban planning. Future research should refine classification accuracy and explore additional clustering methods across diverse urban settings.
Event(s)
19th IBPSA Conference on Building Simulation, BS 2025, 24 August 2025 - 27 August 2025, Brisbane
Publisher
International Building Performance Simulation Association
Type
conference paper