In accordance with the great significance of flood prediction and warning, there has been much research focusing on machine learning models applications (data-driven models) in flood prediction problems. In this paper, we reviewed recent research on the application of machine learning in flood prediction, water-level prediction, discharge prediction, flood depth prediction, etc along with adopted popular indicators for the evaluation of the reliability of machine learning models’ performance. These studies have shown that, unlike traditional numerical models, machine learning models require fewer input parameters, and less simulation time, and do not require extensive knowledge of flood modeling, while still providing good precision prediction results. Besides, the research group has also identified and highlighted some limitations and challenges in the application of machine learning models, along with suggestions for future research orientations to optimize machine learning models in flood prediction.