Abstract
Short-term rainfall prediction is a crucial and practical research area, with the accuracy of rainfall prediction, particularly for heavy rainfall, significantly impacting people's lives, property, and even their safety. Deep learning and RNN cyclic convolutional networks have emerged as important research avenues to address this issue; however, each approach has its limitations. This article integrates their respective key advantages and further optimizes them from the perspectives of model framework and training loss function. Finally, we demonstrate the effectiveness of our optimization measures through experiments.