Testing machine learning models for heuristic building damage assessment applied to the Italian Database of Observed Damage (DaDO)

• DOI: 10.5194/nhess-23-3199-2023
• Publication Date: 2023-10-05T08:27:31Z
• Authors: Subash Ghimire; Philippe Guéguen; Adrien Pothon; Danijel Schorlemmer
• Keywords: Gradient boosting; Boosting

Abstract. Assessing or forecasting seismic damage to buildings is an essential issue for earthquake disaster management. In this study, we explore the efficacy of several machine learning models characterization, trained and tested on database observed after Italian earthquakes (the Database Observed Damage – DaDO). Six were considered: regression- classification-based models, each using random forest, gradient boosting, extreme boosting. The structural features considered divided into two groups: all provided by DaDO only those be most reliable easiest collect (age, number storeys, floor area, building height). Macroseismic intensity was also included as input feature. per determined according EMS-98 scale seven significant occurring in regions. results showed that boosting classification statistically efficient method, particularly when considering basic grouping traffic-light-based system used; example, during post-disaster period (green, yellow, red), 68 % correctly classified. obtained machine-learning-based heuristic model assessment are same order accuracy (error values less than 17 %) traditional RISK-UE method. Finally, analysis found importance with respect conditioned level considered.