Wanigarathna, NadeeshaniXie, YingHenjewele, ChristianMorga, MariantoniettaJones, Keith2025-01-092025-01-092025Wanigarathna N, Xie Y, Henjewele C, et al., (2024) Machine learning application to disaster damage repair cost modelling of residential buildings. Construction Management and Economics, Available online 8 December 20240144-6193https://doi.org/10.1080/01446193.2024.2419413https://dspace.lib.cranfield.ac.uk/handle/1826/23328Data used for this research is publicly available in the following portal. https://openricostruzione.regione.emilia-romagna.it/ricostruzione-privataRestoring residential buildings following earthquake damage requires a significant level of resources. Being able to predict these resource requirements in advance and accurately improves the effectiveness of disaster preparedness and subsequent recovery activities. This research explored how the latest ML algorithms could be used for antecedent earthquake loss modelling. A cost database for repairing residential buildings damaged by the Emilia Romagna earthquake in Italy was analysed using six state-of-the-art ML models to explore their ability to predict repair cost rates(cost per floor area) for a domestic building damaged by earthquakes. A Gradient Boost Regression model outperformed five other models in predicting earthquake damage repair cost rate. The performance of this model was significantly accurate and covers about 76% of the cases. A further SHAP analysis revealed that operational level, damage level and non-housing area of the buildings as top 3 important features when predicting the resultant damage repair cost rate. Overall this research advanced antecedent earthquake loss modelling approaches to increase the accuracy of estimates by incorporating more variables than the widely used damage level based simple methodology.pp. 1-21enAttribution 4.0 Internationalhttp://creativecommons.org/licenses/by/4.0/Earthquakecost modellingmachine learningdamage repair costsdisaster preparedness4005 Civil Engineering40 Engineering33 Built Environment and DesignMachine Learning and Artificial IntelligenceBuilding & Construction33 Built environment and design38 Economics40 EngineeringArticle1466-433X561019ahead-of-printahead-of-print