There are various natural calamities that occur, but floods are particularly crucial due to their widespread impact and devastating effects on communities, infrastructure, and the environment. Effective prediction and timely response are essential to minimize damage and loss of life during such events. Flood prediction is a critical issue in the current decade, and accurate forecasting is essential for mitigation efforts. Due to the bad weather condition it is very difficult for finding the ground data in appropriate time for machine learning to learn effectively. In this study, we address this problem using a machine learning model, specifically a deep neural network using Matlab and Synthetic Aperture Radar images (SAR). SAR have the capability to penetrate bad weather condition. The model was trained using a few-shot learning pipeline, leveraging only 0.6% of the DFC-2024 challenge dataset. Notwithstanding the limited training data, it gained an accuracy of 81.62% and an F1 score of 0.84—defeating existing models trained on the same dataset. These results emphasize the model’s agility and persuasiveness in flood forecast tasks.