In order to establish an insulin evaluation model, this paper studies the insulin evaluation model by presenting a method based on BP neural network. Firstly, a total of 100 cases containing samples of patients with type 2 diabetes and normal glucose tolerance were used as research data. Secondly, a BP neural network model was built, and the physiological parameters of the samples were input into the network for training, thereby obtaining the insulin β cell function index and insulin resistance index. Finally, the insulin β cell function index and insulin resistance index were predicted. The experimental results show that the predicted value of human insulin evaluation model based on BP neural network has a good correlation with the true value, and the correlation of islet β cell function index has reached nearly 90%, which shows that the model can be very good to evaluate insulin secretion.