View in EDS

AI-Driven Database Development for Seismic Loss Estimation in Data-Deficient Regions Using Open-Source Data.

Saved in:
Bibliographic Details
Title: AI-Driven Database Development for Seismic Loss Estimation in Data-Deficient Regions Using Open-Source Data.
Authors: Lee, Kieun1 (AUTHOR) dlrldms0803@ajou.ac.kr, Kim, Taeyong2 (AUTHOR) tyong.kim0117@gmail.com, Yoo, Doo-Yeol3 (AUTHOR) dyyoo@yonsei.ac.kr, Moon, Sungkon4 (AUTHOR) skmoon@ajou.ac.kr
Source: Journal of Computing in Civil Engineering. May2026, Vol. 40 Issue 3, p1-18. 18p.
Subjects: Boosting algorithms, Open data movement, Data extraction, Earthquake hazard analysis, Data libraries
Geographic Terms: South Korea
Abstract: Establishing a building database for city-level earthquake simulations remains challenging due to missing building information. To address this issue, this study proposes an integrated framework that systematically and efficiently collects missing data using artificial intelligence (AI) technologies and multiple open-source data sets. The framework was validated using 222 buildings in the Dasan-dong area of Gyeonggi Province, Korea. To implement the framework, 568,866 data points were gathered from various open-source platforms. The building area and construction year were measured using Mask R-CNN while building height and the number of floors were estimated using a vertical-edge–based method. Structural types and building usage were classified using eXtreme Gradient Boosting (XGBoost), thereby constructing a comprehensive seismic building database. Validation confirmed the framework's robustness: Mask R-CNN detected 99.1% of footprints, with 88.6% of areas and 97.7% of height estimates within 20% relative error and exact floor counts for 84.5% of buildings; XGBoost achieved macro F1-scores of 0.945 for usage and 0.697 for structural type; and R2D earthquake simulations (Mw 5–8) based on the generated database deviated from ground-truth total repair cost loss ratios by only 2.72% on average relative error. The proposed framework effectively supplements missing building information at the city level, providing a crucial foundation for disaster preparedness and enhancing urban resilience. [ABSTRACT FROM AUTHOR]
Copyright of Journal of Computing in Civil Engineering is the property of American Society of Civil Engineers and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:Establishing a building database for city-level earthquake simulations remains challenging due to missing building information. To address this issue, this study proposes an integrated framework that systematically and efficiently collects missing data using artificial intelligence (AI) technologies and multiple open-source data sets. The framework was validated using 222 buildings in the Dasan-dong area of Gyeonggi Province, Korea. To implement the framework, 568,866 data points were gathered from various open-source platforms. The building area and construction year were measured using Mask R-CNN while building height and the number of floors were estimated using a vertical-edge–based method. Structural types and building usage were classified using eXtreme Gradient Boosting (XGBoost), thereby constructing a comprehensive seismic building database. Validation confirmed the framework's robustness: Mask R-CNN detected 99.1% of footprints, with 88.6% of areas and 97.7% of height estimates within 20% relative error and exact floor counts for 84.5% of buildings; XGBoost achieved macro F1-scores of 0.945 for usage and 0.697 for structural type; and R2D earthquake simulations (Mw 5–8) based on the generated database deviated from ground-truth total repair cost loss ratios by only 2.72% on average relative error. The proposed framework effectively supplements missing building information at the city level, providing a crucial foundation for disaster preparedness and enhancing urban resilience. [ABSTRACT FROM AUTHOR]
ISSN:08873801
DOI:10.1061/JCCEE5.CPENG-7077