Floods are among the most destructive natural disasters, leading to severe loss of life, property damage, and economic disruption. Traditional flood prediction methods, which rely heavily on historical data, often lack adaptability to evolving environmental conditions. This article explores the application of machine learning (ML) techniques to enhance flood prediction and management. We investigate the use of supervised learning algorithms such as decision trees, random forests, and other regression methods to predict flood events from meteorological and hydrological data, as well as unsupervised learning methods for detecting anomalies that may indicate impending floods. We propose a comprehensive flood management framework that integrates ML-based prediction models with real-time data from sensors and IoT devices. This framework focuses on early warning systems, risk assessment, and efficient resource allocation. The proposed ML model significantly outperforms traditional methods in predicting flood occurrences, intensity, and impact areas, leading to improved disaster preparedness and reduced economic losses. Subjective and objective evaluations showed that the proposed method could predict flood episodes in certain locations and time periods. Benchmarking trials found that the system configured with random forest provided the best successful prediction, with MSE, MAE, RMSE, and R2 values of 0.7921, 0.9433, 0.9562, and 0.0186, respectively, using synthetic data. Overall, the use of ML constitutes a big step toward creating more robust and adaptable flood management systems.