Improvement of rapid risk analysis in reinforced concrete buildings with machine learning techniques

This thesis focuses on integrating rapid assessment methods for earthquake risk in reinforced concrete buildings with machine learning techniques. Given the devastating impacts and extensive socio- economic costs of earthquakes, the research aims to demonstrate how artificial intelligence and data science methods can serve as effective alternatives to conventional evaluation approaches in accurately identifying high-risk structures. The primary objective is to expedite earthquake preparedness processes, enhance intervention strategies, and reduce costs through a scientifically grounded approach. The hypothesis posits that the Random Forest algorithm will achieve high accuracy and reliability in detecting the seismic vulnerability of reinforced concrete buildings, outperforming traditional methods. Field data collected during 2021 as part of the UDAP project—comprising building identity details, construction year, number of floors, and structural irregularities—was used for analysis. A balanced sample was derived from a larger dataset to train and validate the model. Data preprocessing and hyperparameter optimization techniques were applied to refine the Random Forest model, which demonstrated impressive success rates during both training and testing phases. The findings underscore the model's applicability in real-world scenarios and its potential to contribute significantly to the development of more effective earthquake risk mitigation strategies.