Where to Construct and How to Design Footbridges: A machine learning decision-support tool
43rd Conference on Education and Research in Computer Aided Architectural Design in Europe, eCAADe 2025, Ankara, Türkiye, 1 - 05 Eylül 2025, cilt.1, ss.307-316, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 1
- Basıldığı Şehir: Ankara
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.307-316
- Anahtar Kelimeler: Design Guideline, Footbridge, Prediction Model
- Orta Doğu Teknik Üniversitesi Adresli: Evet
Özet
This study proposes a decision-support tool for footbridge planning by addressing two key questions: where to construct and how to design footbridges. Combining quantitative and qualitative methods, the tool comprises two complementary stages: a prediction model and a design guideline. First, built environment features were explored and collected from locations with and without footbridges in Ankara. Later, three tree-based models (Decision Tree, Random Forest, XGBoost) were trained to predict the need for footbridges at given sites, addressing the “where” question. The model performances were tested through data type transformations and different train-test splits. To provide robust and reliable predictions, the role of randomness was not overlooked, and the top four consistently predicted locations across different data splits for all the models were obtained. XGBoost showed higher performance in traditional terms; however, all the model outcomes were consistent with real-world patterns, as the dataset shows high relevance to the problem domain. The second stage introduced a human-centric approach to address the “how” question by a design guideline developed by synthesizing field observations, pedestrian expectations, and Türkiye’s regulatory criteria, addressing the “how” question. Ultimately, the “where” and “how” were asked together to decide whether to construct a footbridge at a given site, and model predictions were reviewed in context by the authors using the design guideline. By incorporating machine learning, real-world data, and human-centric perspectives, this study offers a comprehensive understanding of footbridge planning.