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

Assessing or forecasting seismic damage to buildings is an essential issue for earthquake disaster management. In this framework, we tested six machine learning models on the DaDO database of observed damage from Italian earthquakes to assess their efficacy in characterizing seismic damage. The models included random forest, gradient boosting, and extreme gradient boosting. The input features were all or a subset of the structural features provided by DaDO, as well as macroseismic intensity. Extreme gradient boosting classification performed best, especially when using basic structural features (age, number of storeys, floor area, building height) and grouping damage according to the traffic-light-based system used, for example, during the post-disaster period (green, yellow, and red). The machine learning-based model had similar efficacy to the traditional Risk-UE method. Finally, the importance of structural features varied depending on the level of damage considered.
The 28th IUGG General Assembly (IUGG2023) (Berlin 2023)