Research On Network Traffic Prediction Model Based On Mode Decomposition

As a research hotspot in the field of network management,network traffic prediction has always attracted the attention of researchers.Traditional single prediction model cannot accurately characterize the self-similarity and long-range correlation of complex network traffic.Decomposition,reconstruction and combined forecasting have become the new direction of network traffic forecasting.This thesis studies the combination prediction model of network traffic based on mode decomposition.The main contents are as follows:1.Aiming at the problem that empirical mode decomposition obtains multiple subsequences and separate prediction results in large computational complexity,a prediction model of real-time network traffic based on empirical mode decomposition and clustering is proposed.Firstly,network traffic is decomposed into simple and relatively stable subsequences using empirical mode decomposition.Secondly,subsequences with similar complexity are clustered together to reduce the number of predictors using improved K-means algorithm.Then,new subsequences are predicted respectively by realtime and recursive kalman filtering models.Finally,the predicted values of all new subsequences are composed.The experimental results show that the model ensures realtime performance,improving the prediction accuracy of traffic.2.Aiming at the problem of mode mixing in empirical mode decomposition and the influence of noise on prediction accuracy,a combination prediction model of network traffic based on ensemble empirical mode decomposition is proposed.Firstly,network traffic is decomposed into single-frequency subsequences using ensemble empirical mode decomposition,which reduces the effect of mode mixing on prediction accuracy.Secondly,the subsequences containing more noise is selected through the energy of the autocorrelation function,and the denoising process is performed.Then,the stationarity of each subsequence is tested.Elman neural network is used to predict the non-stationary subsequences.Autoregressive moving average model is used to predict stationary subsequences.Finally,the predicted values of all subsequences are composed.The experimental results show that the model is superior to other comparison models in mean square error,mean absolute error and accuracy of trend predication,further improving the prediction accuracy of network traffic.