A network traffic prediction method based on empirical mode decomposition and time series self-similar analysis is proposed. Firstly, network traffic time-series high and low frequency components are generated by empirical mode decomposition. Then the component time series is analyzed to determine that the least squares support vector machine model optimized by using the improved harmony search algorithm is used for high frequency components modeling, and the auto regressive integrated moving average model is used for low frequency components modeling and remaining component modeling. Finally, the final prediction result is obtained by RBF neural network nonlinear superposition. Simulation results show that the proposed method has better prediction results and higher prediction accuracy.