This paper provides a comprehensive review of machine learning (ML) models used for flood prediction, highlighting their effectiveness in both short-term and long-term forecasting. Floods are among the most destructive natural disasters, and accurate prediction is crucial for risk reduction, policy development, and minimizing damage. Traditional models, such as physically based and statistical models, have limitations in short-term prediction due to their reliance on complex physical processes and data requirements. ML models, which can learn patterns from historical data without requiring explicit knowledge of physical processes, have emerged as a promising alternative. They offer faster processing, better accuracy, and adaptability to various flood scenarios. The review discusses various ML methods, including artificial neural networks (ANNs), multilayer perceptrons (MLPs), adaptive neuro-fuzzy inference systems (ANFIS), wavelet neural networks (WNNs), support vector machines (SVMs), decision trees (DTs), and random forests (RF). These models are evaluated based on their performance in flood prediction tasks, with a focus on accuracy, speed, and robustness. Hybrid models, which combine multiple ML techniques, are also explored as effective strategies for improving prediction accuracy. The paper emphasizes the importance of data quality and preprocessing in ML-based flood prediction. It highlights the use of various data sources, including rainfall, streamflow, and remote sensing data, to train ML models. The review also discusses the challenges associated with ML models, such as the need for large datasets, computational costs, and the potential for overfitting. Despite these challenges, ML models have shown significant improvements in flood prediction compared to traditional methods, particularly in short-term forecasting. The study concludes that ML models, especially hybrid and ensemble methods, are effective in flood prediction and offer a flexible and adaptable solution for managing flood risks. The review provides insights into the most suitable ML models for different flood prediction tasks and highlights the need for further research to improve the accuracy and efficiency of these models. Overall, the paper serves as a guide for hydrologists and climate scientists in selecting appropriate ML methods for flood prediction based on specific requirements and data availability.This paper provides a comprehensive review of machine learning (ML) models used for flood prediction, highlighting their effectiveness in both short-term and long-term forecasting. Floods are among the most destructive natural disasters, and accurate prediction is crucial for risk reduction, policy development, and minimizing damage. Traditional models, such as physically based and statistical models, have limitations in short-term prediction due to their reliance on complex physical processes and data requirements. ML models, which can learn patterns from historical data without requiring explicit knowledge of physical processes, have emerged as a promising alternative. They offer faster processing, better accuracy, and adaptability to various flood scenarios. The review discusses various ML methods, including artificial neural networks (ANNs), multilayer perceptrons (MLPs), adaptive neuro-fuzzy inference systems (ANFIS), wavelet neural networks (WNNs), support vector machines (SVMs), decision trees (DTs), and random forests (RF). These models are evaluated based on their performance in flood prediction tasks, with a focus on accuracy, speed, and robustness. Hybrid models, which combine multiple ML techniques, are also explored as effective strategies for improving prediction accuracy. The paper emphasizes the importance of data quality and preprocessing in ML-based flood prediction. It highlights the use of various data sources, including rainfall, streamflow, and remote sensing data, to train ML models. The review also discusses the challenges associated with ML models, such as the need for large datasets, computational costs, and the potential for overfitting. Despite these challenges, ML models have shown significant improvements in flood prediction compared to traditional methods, particularly in short-term forecasting. The study concludes that ML models, especially hybrid and ensemble methods, are effective in flood prediction and offer a flexible and adaptable solution for managing flood risks. The review provides insights into the most suitable ML models for different flood prediction tasks and highlights the need for further research to improve the accuracy and efficiency of these models. Overall, the paper serves as a guide for hydrologists and climate scientists in selecting appropriate ML methods for flood prediction based on specific requirements and data availability.