Abstract:Rapid casualty assessments during earthquake disasters are crucial for emergency response. Typically, various factors, such as regional geographic environments, population densities, and building structures, have pivotal impacts on earthquake casualties. We adopted a zonal approach to assess earthquake-related casualties in Mainland China. Specifically, to comprehensively consider the differential impacts of earthquakes on diverse regions, the Chinese Mainland was divided into three zones—northwest, southwest, and east—based on population densities, geographical environments, and building structures. Additionally, the samples were further classified based on the maximum earthquake intensities recorded in these regions. Subsequently, employing the random forest method and bootstrap sampling technique, three parameters—earthquake magnitude, seismic area, and population density—were selected based on the importance of each feature. Thereafter, a particle swarm optimization-extreme learning machine (PSO-ELM) model was established for earthquake casualty assessments. Results indicated that the proposed model demonstrated excellent predictive performance, with good applicability and generalization across diverse regions and intensities, thereby offering valuable technical support for earthquake emergency response efforts and seismic risk assessments.
