Wetlands are vital ecosystems that provide numerous environmental services, such as flood prediction, water purification, and conservation of biodiversity. However, their vulnerability to flooding due to changes in climate and land use requires effective monitoring and prediction models. This study proposes a novel approach to classifying wetland types and predicting flood probabilities based on wetland types using the Efficient Net-ViT model. The model takes advantage of the spatial and temporal characteristics of wetland areas and integrates multiple attention mechanisms to enhance the extraction and learning of features. The performance of the proposed model is evaluated using remote sensing data from satellite imagery and hydrological data, demonstrating its effectiveness in classifying wetland types and predicting flood probabilities with an accuracy of 93.2%. The proposed method provides valuable information for flood management and policy making.