Artificial neural network approaches for disaster management: A literature review

• Published in: International journal of disaster risk reduction 2022-10, Vol.81, p.103276, Article 103276
• Main Authors: Guha, Sreeparna, Jana, Rabin K., Sanyal, Manas K.
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Deep learning; Disaster management; Review
• ISSN: 2212-4209; 2212-4209
• DOI: 10.1016/j.ijdrr.2022.103276

Disaster management (DM) is one of the leading fields that deal with the humanitarian aspects of emergencies. The field has attracted researchers because of its ever-increasing need to find newer and more efficient ways of managing disaster situations to reduce human suffering. This paper reviews 128 articles published in various journals and conference proceedings on Artificial Neural Networks (ANN), a part of Deep Learning (DL), applications in DM from 2010 to 2021. We try to identify the reasons for the superior performance of ANN-based techniques over other techniques. We also classify the extant literature according to applications in different phases and types of disasters. The phases are ‘Mitigation and Preparedness', and ‘Response and Recovery’. The type of disasters includes flood, earthquake, storm, fire hazard/wildfire, and others. We identify some important patterns from this review. The findings establish the following: (i) ANNs are popularly used to predict and manage floods, (ii) Backpropagation Neural Networks (BPNN) are the commonly used architecture, and (iii) Convolutional Neural Networks (CNN) are most promising for extracting social media information during emergencies. We identify the limitations of this study and offer several potential directions for future research.