In the study of vegetable price forecast, as the price is subject to various uncertain factors (weather, supply and demand, etc.), it has attributes such as high nonlinear, randomness and high noise, which would lead to the difficulty in forecasting. But grasping the law of price development and understanding the development trend of price, would help farmers grow the vegetable reasonably, and reduce unbalanced supply and demand. Therefore, we will make use of the characteristics of neural networks such as self-adapt,self-study and high fault tolerance, to build up the model of BP neural network with the training function of L-M for forecasting the vegetable prices. Finally, numerical example proves that the method is effective.