A comprehensive evaluation model of soil nutrient based on BP neural network was set up, which contained 5 cryptic layer nodes and 3 layers. The rank classification criterion of soil nutrient’s assessment indicator system was chosen as the training sample and the test sample of the model. The training and the test were made by using Matlab software. The comprehensive evaluation of farmland soil nutrients was conducted in Ansai County. The results showed that simulation result of BP neural network was consistent with the expected result; the simulation result of BP neural network, fuzzy pattern and principal component analysis were exactly the same. 