Machine learning network suitable for accurate rapid seismic risk estimation of masonry building stocks


COŞKUN O., ALDEMİR A.

NATURAL HAZARDS, 2022 (Peer-Reviewed Journal) identifier identifier

  • Publication Type: Article / Article
  • Publication Date: 2022
  • Doi Number: 10.1007/s11069-022-05553-y
  • Journal Name: NATURAL HAZARDS
  • Journal Indexes: Science Citation Index Expanded, Scopus, IBZ Online, PASCAL, Aerospace Database, Aquatic Science & Fisheries Abstracts (ASFA), CAB Abstracts, Communication Abstracts, Environment Index, Geobase, INSPEC, Metadex, PAIS International, Pollution Abstracts, Veterinary Science Database, DIALNET, Civil Engineering Abstracts
  • Keywords: Seismic risk estimations, Masonry structures, Machine learning, Seismic risk classification, PERFORMANCE ASSESSMENT, VULNERABILITY

Abstract

Most losses from earthquakes are associated with fully collapsed buildings. So, determining the seismic risk of buildings is essential for building occupants in active earthquake zones. Unfortunately, current methods used to estimate the risk state of large building stocks are insufficient for reliable, fast, and accurate decision-making. In addition, the risk classifications of buildings after major natural disasters depend entirely on the experience of the technical team of engineers. Therefore, the decision on risk distributions of building stocks before and after hazards requires more sustainable and accurate methods that include other means of technological advancement. In this study, the building characteristics dominating the seismic risk outcome were determined using a database of 543 masonry buildings. Later, for the first time in the literature, a new, fast and accurate seismic evaluation method is proposed. The proposed method is thoroughly associated with detailed evaluation results of structures with the help of machine learning algorithms. This study utilized an approach in which six machine learning algorithms work together (i.e., Logistic Regression, Decision Tree, Random Forest, K-Mean Clustering, Support Vector Machine, and Ensemble Learning Method). As a result of the analysis of these algorithms, the correct prediction rates for the learning database (i.e., 434 buildings) and the test database (i.e., 109 buildings) of the proposed method were determined as approximately 96.67% and 95%, respectively. Lastly, machine learning algorithms trained by structures with known after seismic risk results are developed. The proposed method managed to classify risk states with the accuracy of 84.6%.