Due to the impact of global climate change and human activities on ecological environment, mountain flood has become one of the most serious natural disasters in China. Flash flood forecast is the key link to prevent and reduce the loss of flash flood disaster, which is of great significance to protect people’s life and property safety. In order to improve the accuracy and timeliness of the forecast, this paper adopts big data and artificial intelligence technology to monitor the characteristics of mountain flood disaster and the dynamics before the occurrence, estimate the loss, and sort out the data and submit it to the government for warning. This paper makes an in-depth study of small and medium-sized rivers, borrows small experience, and applies it to a larger field. This paper uses big data mining, machine learning, deep learning and other technologies to build an intelligent early warning system for flash flood forecast of small and medium-sized rivers. In addition, this paper selects the historical hydrometeorological data of small and medium-sized rivers in Sichuan, Yunnan, and Guizhou provinces in China to compare and analyze the experimental simulation values to ensure the richness and authenticity of the data. The experimental results show that the predicted runoff depth is 16.1 mm and 68.6 mm. Compared with the actual runoff depth, the error values are 28.5% and 19.9%, respectively.