LRD network traffic predicting based on SRD model

Abstract

The prediction of long range dependence (LRD) is the critical problem in network traffic. The traditional algorithms, such as Markov model and ON/OFF model, may provide high computation cost and low precision. In this study, a novel method based on empirical mode decomposition (EMD) and ARMA model was proposed. The researchers adopted EMD to decompose the network traffic data which would be decomposed into several IMF (Intrinsic Mode Function) components and found that those IMF components had no longer self-similar property. Experiment results show that EMD could offer the function of canceling the LRD in traffic data. After transforming LRD to SRD (short range dependence) by EMD processing, the LRD traffic data could be predicted with high accuracy and low complexity by ARMA model. Meanwhile, the results indicate the usefulness of EMD in the applications of network traffic prediction.