Product sales prediction of e-commerce enterprises is an important basis for strategic decision-making. Previous studies have primarily used the Back Propagation (BP) neural network and regression analysis to predict sales. The BP neural network has disadvantages such as falling into local minima and slow convergence rate, while regression analysis can not fit nonlinear data well. The radial basis function (RBF) neural network can effectively overcome the above shortcomings. This paper adopts the RBF neural network to predict sales. Since sales prediction is affected by various factors, this paper includes various influential factors into the research category. In this paper, RBF neural network and BP neural network are trained and tested successively, and a regression analysis model is constructed. The conclusion shows that the average absolute error, root mean square error, and average absolute percentage error of the RBF neural network are 3/8, 1/2 and 1/3 of the corresponding values of the BP neural networks, and 5/8, 3/4 and 4/7 of the corresponding values of regression analysis. Therefore, the RBF neural network has a higher prediction accuracy and can better predict product sales of e-commerce enterprises.