Interpretable machine learning and feature selection for seismic risk prediction of low-rise reinforced concrete buildings
BULLETIN OF EARTHQUAKE ENGINEERING, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1007/s10518-026-02667-7
- Dergi Adı: BULLETIN OF EARTHQUAKE ENGINEERING
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, Geobase, INSPEC, Academic Search Ultimate (EBSCO), Natural Science Collection (ProQuest), Earth, Atmospheric, & Aquatic Science Collection (ProQuest), Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
- Erciyes Üniversitesi Adresli: Evet
Özet
This study evaluates the performance of feature importance and feature selection methods for rapidly and reliably estimating the seismic risk levels of reinforced concrete (RC) buildings prior to an earthquake. The aim is to identify the most influential parameters in machine learning models and to investigate whether accurate predictions can be achieved using a reduced set of input features. A simulation-based dataset representing an existing building stock was utilized. Analyses using three machine learning models demonstrate that combining feature importance and feature selection improves prediction performance. LightGBM with normalized feature importance and CORR-based feature selection achieved the highest performance, reaching a macro-F1 score of approximately 0.91 while using fewer input parameters and outperforming models trained with the full feature set. The findings demonstrate that reliable seismic risk predictions can be achieved using a limited number of critical parameters, thereby enabling rapid and efficient large-scale building stock assessments. In addition, the proposed framework provides an interpretable and data-driven decision-support approach for regional seismic risk evaluation and urban-scale vulnerability screening applications.