Building A Deep Learning Model for Multi-Label Classification of Natural Disasters
Qiang Cao, Yan Liu, Guangxu Wang, Yuxin He, Kuanglan Wang, Stephen Shaoyi Liao, Lixin Pu · 2023
Natural disasters, such as earthquakes, hurricanes/typhoons and wildfires, usually cause severe damage. Disaster response and management is a great challenge to the authority. Current studies usually focus on a single disaster identification using social media data. In reality, there are relationships among different types of disasters. And several disasters may happen simultaneously. In this study, we explore the role of the deep learning model in multi-label disaster classification. We build a deep CNN model for multi-label classification with the instruction of a high-order strategy. We train and validate our model using a professional low-altitude disaster dataset, LADI. We find our proposed deep learning model with the transfer learning method outperforms many other machine learning models in the previous study.