F.LO.O.D.S.A.F.E: Flood Localization and Observation for Optimal Detection and Strategic Alert Forecasting Engine

Floods are among the most devastating natural disasters, causing economic and environmental damage. This paper presents a machine-learning-based flood prediction and alerting system that integrates deep learning techniques with satellite and weather data. The system utilizes convolutional neural networks (CNNs) for image classification and U-Net architectures for precise flood mapping. Random forests and gradient boosting machines analyse real-time weather data to predict flood risks. Additionally, a dynamic evacuation route planning system provides optimized routes based on real-time conditions. The integration of cloud platforms ensures scalability and real-time response, making F.L.O.O.D.S.A.F.E a viable solution for flood-prone regions.