Disaster identification scheme based on federated learning and Cognitive Internet of Vehicles

The rate of disaster occurrences has been increasing over the last decade due to the alarming effects of global warming. A major challenge with such disasters is identifying their nature before substantial loss of lives and property occurs. Existing systems often fail to determine the type of disaster until significant damage has been done. This research proposes a novel scheme to identify disasters as they occur, leveraging Federated Learning (FL) and Cognitive Internet of Vehicles (CIoV) since vehicles are a common presence in disaster scenarios. The proposed scheme utilizes various machine learning (ML) and deep learning algorithms to predict disaster types in real-time. Additionally, it introduces a custom federated averaging algorithm to maintain result privacy. The research evaluated the scheme’s performance using a data set of recorded disasters from various countries, training different algorithms to determine optimal results. The results indicate that the proposed scheme can achieve a 90% accuracy in disaster-type identification using deep learning and random forest algorithms.