Abstract: The traffic with data packet transmission in various network conditions exhibits convincingly self-similarity causing the long range burstiness which cannot be captured by traditional telecommunication traffic models based on Poisson process or Markov process. The updated explicit high-resolution measurement and researches for the traffic reveal that the heavy tailness existing extensively in the network brings about the self-similarity of the traffic. The information extraction from the self-similarity and long range dependence is the key fact for the exact prediction of the long range bursty traffic. Two distinctive AutoRegressive predictors based on c}stable self-similar traffic model were presented. The predictors including FAR(Fractional AutoRegressive) ,FNAR(Fractional Nonlinear AutoRegressive) can minimite the dispersion according to the criteria with infinite variance. The final predicted values with the different schemes were obtained by combining the previous two individual predicted values for the higher predicted precision.

Key words: Long range dependence, Traffic, Autoregressive, Prediction