This article discusses the use of machine learning techniques, specifically extremely randomized trees, for dam management and diagnosis based on observation data. The study found that the water level in the reservoir, rainfall, and temporal gradient of water level are important factors for accurate prediction.