Aiming to the dynamics and nonlinearities of stock price, a stock price prediction model that based on support vector regression (SVR) with parameters optimized by improved genetic algorithm (GA) was proposed. First, the wavelet was used to de-noise the samples of stock price. Then the SVR model whose parameters were optimized by improved GA was utilized to predict and assess the data de-noised by wavelet. The result demonstrated that the improved wavelet-GA-SVR model has good prediction effect, and it is significant to the study of the prediction of stock price.