Aiming at the problem that the prediction model of small sample product is not accurate enough, a prediction method of product performance parameters is proposed. Based on the learning algorithm of radial basis function (RBF) neural network, the idea of transfer learning is added, and the historical test data of small sample products and the test data of other products with the same model and batch are fully learned as the source domain knowledge, which makes up for the problem of poor prediction accuracy of product performance parameters caused by the lack of labeled sample data in the current field. The results show that the prediction accuracy of this method is high.