Machine Learning and the Economic Losses in Disasters: Progress and Future Trends

Climate change-induced disasters pose a significant global societal challenge, resulting in disruption and substantial economic losses. To combat such challenges, it is crucial to develop resilience and enhance the ability of any system (e.g., critical infrastructure, cities, or communities) to handle unforeseen events. Different approaches have been employed to support decision-makers in performing resilience assessments. For example, disaster loss estimation is an essential process for determining associated losses and supporting practical actions. However, these methods are resource- and time-consuming due to the data used. Machine learning (ML) techniques have emerged as powerful tools to overcome the challenge of analyzing vast amounts of data. Despite significant strides in applying ML techniques, more comprehensive reviews must be done on methodological development and ML applications in disaster loss estimation. This study aims to review the literature on ML techniques and their application in natural disaster loss estimations. A total of 676 studies were collected, and 56 were selected for detailed analysis. The study presents an overview of the ML techniques and applications for estimating economic losses in disaster scenarios. This review identifies the remaining challenges, limitations, and opportunities. Based on the outcomes of this review, a taxonomy was generated to characterize the various aspects of this area. The findings offer significant insights for policymakers, practitioners, and researchers, filling a critical void in current research.