关键词: 非平稳时间序列, 小波分析, 最小二乘支持向量机, 自回归, 预测

Abstract: For improving the prediction accuracy of network traffic，a new combination prediction model is proposed based on wavelet analysis and AR-LSSVM.The network traffic series are decomposed and reconstructed using Mallat algorithm，a low frequency signal and several high frequency signals are gotten.The high frequency signals with stationary characters are predicted with Auto-Regression（AR） models，and the low frequency with non-stationary character is predicted with Least Square Support Vector Machines（LSSVM）.The final prediction result of the original traffic series is the superimposition of these respective prediction results.The simulation results show that the method has higher prediction accuracy and steady prediction performance.

Key words: non-statlonary time series, wavelet analysis, least squares support vector machines, auto-regression, prediction