An Intelligent Flood Prediction System Using Deep Learning Techniques and Fine Tuned MobileNet Architecture

A flood can cause significant damage and loss of life and economic disruption. Early warning and accurate forecasting of such disasters can help minimize the effects of these natural disasters by helping with evacuation plans and allocation of resources. The main objective of this study was to develop an automatic model that can predict the precipitation seasons. It was done using the ANNs and the CNN structure. The study utilized real-time images including normal and flood-affected scenes. The results of the study revealed that the CNN models performed well in terms of their accuracy, precision, and F1-score. The CNN models have performed well in terms of their accuracy rates with 96.55 classification accuracy for both 10 and 100 epochs. The IFPS could help authorities identify potential flood threats and take immediate action to protect their communities. The proposed IFPS was evaluated against existing flood prediction tools. Finally, the performance of IFPS is compared with three deep learning algorithms such as ResNet-50, VGG-16, and Inception V2. The results indicated that the deep learning system was more accurate and faster than the traditional methods and the pre-trained models.