With the development of science and technology, people pay more attention to predicting the price of stock by using machine learning algorithm. Different algorithms have been used to forecast the direction of the stock prices and researchers have already got some achievements. In this paper, we propose a feedforward neural network model and use one year’s data to train it until the accuracy of prediction is more than 90%. Predicting stock prices will be concerned more in the future. And many algorithms will be adopted to improve the prediction accuracy.