Research On Decomposition-based Network Traffic Combination Prediction Method

With the proliferation of smartphones,data traffic is growing exponentially,and to manage this explosive growth,accurate network traffic forecasting is needed to plan network capacity.Therefore,accurate network traffic prediction can help operators respond to impending congestion early with advanced network capacity expansion,adjustment,and optimization.In addition,with the popularity and application of computer technology,critical information related to daily life and work has been transmitted through large-scale digital networks.Network traffic prediction is a reliable way to monitor network communication security.In summary,network traffic prediction is essential for effective network maintenance,load assessment and security alerts.The self-similarity,periodicity,chaos,multi-scale and other characteristics of modern network traffic make it challenging to predict network traffic.In this thesis,a network traffic prediction model(VMD-SCNs-WOA-SVM model,abbreviated as VSWS model)based on Variational Mode Decomposition(VMD)combination with Stochastic Configuration Networks(SCNs)model and Support Vector Machine(SVM)model is proposed.This combination prediction model aims to improve the accuracy and reduce the time complexity of network traffic prediction.First,the VMD decomposition is performed on the network traffic sequence.The decomposed components eliminate the long correlation of the time series and highlight the local features,which can reduce the non-smoothness of the time series and solve the problem that the data itself is difficult to predict.Second,the approximate entropy quantification complexity of each component is calculated.The components obtained by the VMD decomposition have different characteristics.Therefore,choosing a suitable prediction model for each component is essential.In this thesis,the approximate entropy algorithm is introduced to measure the complexity of each component,which is the basis for choosing a prediction model.The components with higher approximate entropy have higher complexity and are predicted using the SCNs model.The components with lower approximate entropy have lower complexity and are predicted using the SVM model.At the same time,the Whale Optimization Algorithm(WOA)is introduced to optimize the parameters of the SVM to improve the performance accuracy of the prediction model.The prediction value of each component superimposes the final network traffic prediction value.Finally,in this thesis,simulation experiments are conducted on two sets of actual collected network traffic data.Root Mean Square Error(RMSE),R~2,Mean Absolute Error(MAE),Mean Absolute Percentile Error(MAPE),relative error plot of prediction results,and box line plot of prediction errors is used to measure the prediction accuracy of the prediction model and to compare the training time and prediction time of the model.The experimental results show that the VSWS model obtains higher prediction accuracy and lower time complexity than other network traffic prediction models.