A Machine Learning Framework for Detecting Morphological Shift in Rapidly Urbanizing Cities: The Case of Ankara, Türkiye

ÖRNEK M. A.

44th Conference on Education and Research in Computer Aided Architectural Design in Europe, eCAADe 2026, Lübeck, Germany, 7 - 11 September 2026, vol.1, pp.429-438, (Full Text)

  • Publication Type: Conference Paper / Full Text
  • Volume: 1
  • Doi Number: 10.52842/conf.ecaade.2026.1.429
  • City: Lübeck
  • Country: Germany
  • Page Numbers: pp.429-438
  • Keywords: Local Density, Machine Learning, Urban Morphology, Urban Sprawl
  • Middle East Technical University Affiliated: Yes

Abstract

This study investigates morphological changes in residential blocks over three decades of rapid urbanization, utilizing a machine-learning methodology to objectively characterize the typological changes and identify the most prominent indicators of new building stock. The study is situated in Çankaya, a district of the city of Ankara, chosen for its neighborhoods with contrasting developmental histories. The dataset consists of 106 buildings with manually collected ground-truth age labels, supplemented by extracted geospatial data through Python-based mining and the OSMnx module. Thirteen morphological features were derived from raw data related to geometry, context, location, network, and density to characterize settlement patterns. Random Forest and Decision Tree classifiers were trained to predict building age across four predefined categories. The Random Forest model achieved an overall accuracy of 63.64% on the test set and 60.48% under cross-validation, performing strongest for the extreme age categories while showing poorer performance in differentiating the intermediate classes. Proximity to the city center emerged as the most important predictor, followed by the fractal dimension and local density measures. These findings quantitatively confirm the observed trends of outward expansion and the process of "de-densification" of Ankara's urban fabric. Beyond basic prediction, the study demonstrates that machine learning can serve as an analytical tool for exploring urban history and addressing data gaps for future city planning.