Application of Machine Learning in Flood Forecast: A Survey

Floods are one of the most dangerous natural disasters, which are highly hard to predict. Climate change and urbanization have increased the frequency or intensity of floods in recent years, and the resulting casualties and economic losses have also increased significantly. With the rapid development of computing power, flood forecast models based on machine learning have gradually emerged. These models that are trained on historical data contain rich information, which is conducive to the analysis and utilization of data. Compared to the traditional physical flood forecasting model, machine-learning-based models can obtain more satisfactory performance. To demonstrate recent advances in flood prediction, this paper presents an overview of recent flood prediction methods using machine learning. We classifier the model based on various strategies and list a variety of recent work in flood prediction. Furthermore, we review the evaluation indexes and compare the pros and cons of different models, hoping to present insights and motivations for future research directions.