Abstract: An assessment system of water resources carrying capacity indicator was constructed according to pressure-state -response （PSR） framework, and the random samples were constructed based on index threshold. On this basis, the random forest algorithm was applied to evaluate the water resources carrying capacity in Hubei province. The results showed that the relationship between the index factors and the water resource carrying capacity could be well fitted by the random forest with a high accuracy. The scheme was of good maneuverability and certain application value. The random forest algorithm can recognize the importance of variable unbiased, and the research showed that the modulus of water demand and the modulus of water supply were the key factors affecting the carrying capacity of water resources in Hubei province. The spatial distribution of water resources carrying capacity in Hubei province was uneven. The carrying capacity of water resources in western Hubei and Xianning areas was of good grade, while that of eastern Hubei and middle Hubei was generalgrade. The Jianghan plain area with Wuhan as the core belonged to poorgrade. It was suggested that regional management of water resources should be strengthened to promote the development of water resources conservation ability and economic society synchronously.

Key words: random forest algorithm, water resources carrying capacity, indicator system, Hubei province