Smart Flood Detection and Communication in Hazardous Zones via MANET-Based IoT and Parallely Distributed Slimmable Neural Networks

Disasters can be mitigated by an early warning signal and proper communication within the hazardous environment using the MANET technology. However, the exact prediction of disaster situation is needed for the timely disaster management. Hence in addition to MANET technology, deep learning algorithms and Inter of Things (IoT) can also be integrated into disaster detection and communication systems. The presented research developed a novel parallely distributed slimmable neural network with orchard algorithm for the accurate flood disaster prediction. The model processed the dataset containing various measurements such as temperature, humidity, precipitation, air, soil moisture and rainfall level that are collected by the IoT sensor deployed in the various location of the hazardous environment. The gathered data are initially pre-processed using the correlation coefficient min-max normalization approach. Further, the relevant characteristics of the flood disaster from the data are extracted through spike driven transformer process. These refined and dimensionality reduced data are then entered into the proposed framework for the disaster prediction. Here the integrated orchard algorithm increased the disaster monitoring accuracy on the basis of better optimized parameters. The model resulted 95% accuracy, 93% precision, 94% recall and 96% f-score. Therefore, the presented method is an effective prediction mechanism for the disaster management in IoT MANET.