The existing assessment models for water-soil environment quality are usually established on the relationships between assessment indicators and their assessment criteria. Such kinds of models are varied with regional scale, and always need a mass of calculation work. This paper tried to find a general assessment model based on a given water-soil quality assessment criteria. In this process, stochastic technology was used to simulate enough assessment indictor series, and then, assessment model was built up by using artificial neural network to assess these series. This model could reduce work load, and needn't construct functional relations between assessment indicators and their criteria and calculate weigh value. A case study in a basin with the highest level of water resources utilization showed that the model was practical and convenient, and could be used in basin water-soil quality assessment.