| dc.description.abstract |
On April 25, 2015, a devastating earthquake of magnitude Mw 7.8 hit Nepal, killing around 9000 people and injuring 22000 more. Following the disaster, extensive field research and inspections were conducted in Nepal to determine the extent of damage to the earthquake-affected structures. The post-earthquake investigation procedure becomes extremely difficult due to the vast number of structures and types of buildings in the area. However, knowing a building’s description beforehand can assist in determining the extent of possible damages due to a large event. In light of this, the present study aims to provide an effective formulation for building vulnerability assessment using several parameters, such as number of floors, construction materials, house type (public or private), and age of building. A huge dataset comprising building information of around 3,50,000 buildings on 39 variables is used for this purpose. Six machine learning methods, namely logistic regression, decision-tree classifier, k-nearest neighbor, linear discriminant analysis, random forest, and extreme gradient boosting algorithms are implemented. Based on the score, the grading boosting algorithm is found to be the most suitable algorithm. The findings are helpful for better urban planning, social policymaking, suitable material identification for building construction, and moreover, to set up a national level disaster risk reduction (DRR) strategy to minimize earthquake losses in Nepal |
en_US |