Because of the nonlinear and uncertain characteristics of the financial time series, it is difficult to obtain satisfactory results by using traditional forecasting methods. This investigation proposes a method to predict the stock price time series based on the LSSVM model which uses PSO algorithm to optimize its parameters. The PSO algorithm is used to optimize the penalty factor and kernel function parameter of LSSVM with fast convergence speed and global convergence ability. This proposed model can improve the prediction accuracy of financial time series and has better generalization ability when using it to predict time series of stock price in financial market compared with the traditional methods.